Tapestry and HTML forms

Transcription

Tapestry in Action
Licensed to Rodrigo Urubatan Ferreira Jardim
[email protected]
Tapestry in Action
HOWARD M. LEWIS SHIP
MANNING
Greenwich
(74° w. long.)
Licensed to Rodrigo Urubatan Ferreira Jardim <[email protected]>
For online information and ordering of this and other Manning books, go to
www.manning.com. The publisher offers discounts on this book when ordered in quantity.
For more information, please contact:
Special Sales Department
Manning Publications Co.
209 Bruce Park Avenue
Fax: (203) 661-9018
Greenwich, CT 06830
email: [email protected]
©2004 by Manning Publications Co. All rights reserved.
No part of this publication may be reproduced, stored in a retrieval system, or transmitted,
in any form or by means electronic, mechanical, photocopying, or otherwise, without
prior written permission of the publisher.
Many of the designations used by manufacturers and sellers to distinguish their products
are claimed as trademarks. Where those designations appear in the book, and Manning
Publications was aware of a trademark claim, the designations have been printed in initial
caps or all caps.
Recognizing the importance of preserving what has been written, it is Manning’s policy
to have the books they publish printed on acid-free paper, and we exert our best efforts
to that end.
Manning Publications Co.
Copyeditor: Liz Welch
209 Bruce Park Avenue
Typesetter: Denis Dalinnik
Greenwich, CT 06830
Cover designer: Leslie Haimes
ISBN 1-932394-11-7
Printed in the United States of America
1 2 3 4 5 6 7 8 9 10 – VHG – 08 07 06 05 04
To my parents, who bought me my first computer when I was 13
and had no idea what they had started
brief contents
PART 1 USING BASIC TAPESTRY COMPONENTS ........................... 1
1
■
Introducing Tapestry
3
2
■
Getting started with Tapestry
3
■
Tapestry and HTML forms 92
4
■
Advanced form components 133
5
■
Form input validation
38
169
PART 2 CREATING TAPESTRY COMPONENTS ............................213
6
■
Creating reusable components
7
■
Tapestry under the hood
8
■
Advanced techniques
215
269
322
PART 3 BUILDING COMPLETE TAPESTRY APPLICATIONS ........381
9
■
Putting it all together
383
10
■
Implementing a Tapestry application
403
vii
viii
BRIEF CONTENTS
APPENDIXES
A
■
Getting involved with Tapestry
479
B
■
Building the examples with Ant
485
C
■
Tapestry component reference 493
D
■
Tapestry specifications
516
contents
foreword xvii
preface xix
acknowledgments xxi
about this book xxiii
about the title xxvii
about the cover illustration
xxviii
PART 1 USING BASIC TAPESTRY COMPONENTS ...................... 1
1
Introducing Tapestry 3
1.1
What are web applications?
1.2
What are Java servlets?
5
7
Understanding servlet multithreading 9 Managing server-side
state 10 Using Struts with servlets 12 Understanding the
limitations of servlets 12
■
■
1.3
■
Why do we need Tapestry?
16
What is a framework? 16 What is a component? 17
What is Tapestry? 18 Comparing Tapestry to Swing 21
■
■
1.4
Understanding Tapestry’s goals
Simplicity 22
Feedback 23
■
Consistency
22
22
■
Efficiency
23
ix
x
CONTENTS
1.5
How does Tapestry work?
25
What’s in a Tapestry application? 27 Tapestry’s ModelView-Controller (MVC) pattern 29 Tapestry classes 33
■
■
2
1.6
Using Spindle
1.7
Summary
35
37
2.1
38
Introducing the Hangman application 39
Determining the application flow 41 Creating page
mockups 42 Defining the domain objects 44
Defining the pages 49
■
■
2.2
Developing the Home page
51
Understanding the Home page specification 52 Rendering the
Home page 54 Defining the Home page class 57
Examining the Visit object 60
■
■
2.3
2.4
Implementing the Home page using standard
servlets 62
Developing the Guess page 65
Displaying the remaining guesses 73 Generating the guessed
word display 79 Selecting guesses 83
■
■
3
2.5
Developing the Win and Lose pages
2.6
Configuring the web.xml deployment descriptor 89
2.7
Summary
91
92
3.1
Understanding HTML forms
3.2
Creating a simple login form 96
Implementing the Login page class
properties 101
3.3
89
93
99
Using specified
■
Understanding the Form component
103
Developing forms without Tapestry 103 Developing forms
with Tapestry 105 Handling form submissions 108
■
■
3.4
Using basic form control components
109
Understanding the essentials 110 The Checkbox
component 111 Radio and RadioGroup components
■
■
111
CONTENTS
Select and Option components
components 118
3.5
Creating a to-do list
113
■
Submit and ImageSubmit
120
Defining the data object 121 Creating the ToDo HTML
template 122 Specifying properties in the page
specification 125 Initializing the toDoList property 126
Handling reordering 127 Deleting completed items 128
Triggering stale links 129
■
■
■
■
3.6
4
Summary
132
Advanced form components
133
4.1
Introducing the advanced form components
4.2
Creating drop-down lists with PropertySelection 134
134
Adding priority levels to the ToDo application 136 Updating
the HTML template 138 Implementing the page class 139
Implementing the model 140 Using enums instead of
integers 144
■
■
■
4.3
Recording data in the form with Hidden
4.4
Looping within a form using ListEdit
Using the ListEditMap
application 154
5
153
■
148
151
Using ListEdit in the ToDo
4.5
Handling file uploads
4.6
Creating pop-up date selections using DatePicker
4.7
Summary
160
165
168
169
5.1
170
Validating user input
Using FieldLabels in conjunction with ValidFields 173
Using validators 173 Using validation delegates 174
Using helper beans 178
■
5.2
Building the Register page 179
Creating the Register HTML template
Register page specification 189
181
■
Creating the
5.3
Validating input based on regular expressions 195
5.4
Customizing label and field decorations 199
5.5
Enabling client-side validation
202
xi
xii
CONTENTS
5.6
Handling form-level validations
5.7
Using validation without ValidField 208
5.8
Summary
205
211
PART 2 CREATING TAPESTRY COMPONENTS ..................... 213
6
215
6.1
Creating simple template components 216
6.2
Creating the component specification
219
Specifying the component’s Java class 219 Discarding the
component’s body 220 Forbidding informal parameters 221
Declaring parameters 222 Reserving names 230
■
■
■
6.3
Creating the Digit and Scaffold components
230
Specifying the digit parameter 232 Using the digit
parameter 232 Creating the template 233 Using the Digit
component 233 Using the Scaffold component 234
■
■
■
■
6.4
Creating the Letter component 234
Specifying the Letter component 235 Implementing the Letter
component 236 Using the Letter component 237
■
■
6.5
Building the Spell component 238
Implementing the Spell component
component 240
6.6
239
Building the Border component
■
Using the Spell
242
Creating the Border template 243 Creating the Border
specification 244 Using the Border component 244
■
■
6.7
Creating interactive, reusable components 246
Introducing the Pet Store image map 246 Specifying the
DirectArea component 247 Implementing the DirectArea
component 248 Using the DirectArea component 252
■
■
■
6.8
Using component libraries 254
Declaring libraries
6.9
254
■
Referencing library components
Packaging components into libraries
256
261
Creating the library specification 262 Creating the library
specification 264 Creating the FormError component 264
Using the FormError component 267
■
■
6.10
Summary
268
CONTENTS
7
269
7.1
Processing requests
7.2
Understanding the application servlet
270
271
Servlet request processing 271 Understanding server-side
state 272 Managing server-side state in a cluster 274
■
■
7.3
Understanding the Tapestry engine
7.4
Understanding engine services
277
279
What’s the problem with application URLs? 279 How does
Tapestry handle application operations? 280 Using the home
service 281 Rendering pages with the page service 283
Linking to listener methods with the direct service 284
Creating bookmarkable links using the external service 290
■
■
■
7.5
Page rendering in detail
291
Using markup writers 292 Going beyond HTML
Understanding the page-rendering sequence 297
Using page-rendering events 297
■
7.6
Loading and pooling pages
295
299
Retrieving pages from the pool 300 Creating new page
instances 302 Returning pages to the pool 304
■
■
7.7
Using persistent page properties
7.8
Using specified properties
7.9
Localizing Tapestry applications
Using Java localization
features 311
7.10
8
Summary
308
■
309
Using Tapestry’s localization
321
Advanced techniques
8.1
310
306
322
Creating new engine services 323
Defining a banner ad system 324 Defining the data
model 326 Accessing the data model as an application
extension 328 Implementing the BannerAd component 329
Implementing the banner service 332 Creating the library
specification 336 Building a banner ad application 337
■
■
■
■
■
8.2
Client-side scripting
339
Defining the CreditCardField component 341 Working with
the Body component 344 Creating the Tapestry script
■
■
xiii
xiv
CONTENTS
specification 345 Creating the CreditCardField
specification 355 Creating the CreditCardField
component 356 Using the component 368
■
■
■
8.3
Integrating with JavaServer Pages
369
Redirecting to a JSP 369 Linking JSPs to Tapestry
pages 371 Submitting JSP forms into Tapestry 375
■
■
8.4
Summary
379
PART 3 BUILDING COMPLETE TAPESTRY APPLICATIONS ...... 381
9
383
9.1
Introducing the Virtual Library
9.2
Performing searches 385
Changing the table sort order
results 387
387
384
■
Paging through the
9.3
Logging in and registering
9.4
Borrowing books 389
9.5
Getting details about books and persons 390
Viewing book details
9.6
390
■
388
Viewing a person
392
Managing your books 393
Editing a book 394 Deleting a book 395 Returning
books 395 Adding a new book 396 Editing your
profile 396 Giving away books 396
■
■
■
■
■
9.7
Administering the Virtual Library
Editing users
books 400
9.8
10
Summary
399
■
399
Editing publishers
400
■
Transferring
402
Implementing a Tapestry application 403
10.1
Looking at the application layers
10.2
Organizing EJB access
404
406
Handling authentication 407 Accessing Enterprise
JavaBeans 407 Tracking user identity with the Visit
object 407 Understanding page inheritance 410
■
■
■
xv
CONTENTS
10.3
Implementing the Search page
413
Identifying application-specific components 415 Referencing
the engine 416 Specifying the page class and properties 417
Performing searches 418
■
■
10.4
Implementing the BookMatches page
419
Handling paging and column sorting 420 Using the Browser
component 425 Executing queries and re-queries 426
■
■
10.5
Implementing the Browser component
430
Specifying Browser’s parameters 430 Getting results from the
BookQuery bean 432 Rendering the Browser
component 433
■
■
10.6
Implementing the ColumnSorter component
434
Creating the ColumnSorter HTML template 434 Specifying
ColumnSorter parameters 435 Responding to the user 437
■
■
10.7
Implementing the Border component
Handling user login
10.8
442
Authenticating the user
■
438
Linking to MyLibrary
443
445
Remembering the user 446 Clearing the password field
Invoking the login operation 449
■
10.9
Creating bookmarkable links
447
449
Creating the BookLink component 450 Displaying the Book
on the ViewBook page 451 Creating the PersonLink
component 453 Displaying the Person 455
■
■
■
10.10
Editing a Book
457
Tracking the Book ID
JavaScript 460
459
■
10.11
Giving books away 463
10.12
Editing the publishers 464
Generating dynamic
Constructing the EditPublishers template 464 Declaring
properties for the EditPublishers page 465 Creating the
ListEditMap 467 Updating the publishers 469
■
■
■
10.13
Editing the list of users
470
Creating the ListEditMap subclass
submission 472
471
■
Handling the form
xvi
CONTENTS
10.14
Creating the web deployment descriptor 474
Deploying web applications as root 476
enterprise application as root 476
10.15
■
Deploying an
Wrapping it all up 477
appendix A: Getting involved with Tapestry 479
appendix B: Building the examples with Ant 485
appendix C: Tapestry component reference
493
appendix D: Tapestry specifications 516
index 537
foreword
My involvement with Tapestry began in the autumn of 2001. I read about the
framework in an article in ONJava magazine. At the same time, our company
was poised to begin several new web projects, and we were looking for a way to
avoid the problems inherent in building complex web pages and forms with
the standard tools. We analyzed a large number of frameworks, but Tapestry
immediately attracted our attention, with its unique development method and
its helpful community.
The first glance was not misleading—Tapestry proved to be a powerful and
helpful instrument in practice as well. The component structure was not simply an add-on but was entrenched in the philosophy of the framework. We
also discovered that Tapestry offered a number of other powerful features that
proved to be critical in our work. For example, it allowed a clean separation
between Java and HTML, and made it possible for the design work on the
application to continue well after the code had been completed—and it could
be performed by designers who never had to know anything about Tapestry. It
provided internationalization capabilities well beyond simply replacing text
with its translation. The framework was designed with EJBs and clustering in
mind, and integrated with them effortlessly.
Today our company has libraries containing hundreds of Tapestry components. Some of these are simple, such as a Login component that manages
authentication using HTTP cookies. Others are far more complex, constructed
xvii
xviii
FOREWORD
from smaller components and full of intricate JavaScript, such as a tool for
dynamically defining web forms. All of these components can simply be taken off
the shelf and plugged into our latest application with ease. We do not have to
worry how the various components will work together—we know that they will do
so by design.
During the two years we have worked with Tapestry, its development has not
stood still. While the earlier versions of the framework concentrated on delivering power, version 3.0 (described in this book) concentrates on both delivering
maximum ease of use for programmers and enhancing their productivity. It lowers the entry requirements, decreases the amount of developer effort needed,
and makes the framework as easy to use as a scripting language for simple applications. The result is that you can achieve the same results as before but with
much less coding.
The majority of these improvements have occurred and moved forward due
to feedback and ideas from Tapestry users and contributors. The active community surrounding the framework helps it remain focused on resolving the problems developers encounter in the real world. A major commendation for
encouraging and integrating the contributions must also go to Howard Lewis
Ship, the author of this book. As the creator and original designer of Tapestry,
he has made a remarkable effort to listen to users, understand their needs, and
address their requirements. The other Tapestry contributors have followed his
lead and have extended the framework to provide support for a variety of new
functionality areas. The community has also been instrumental in the process
that made Tapestry a part of the Apache Jakarta family. With this step, it became
a companion project of other popular projects, including Struts, Tomcat, and
log4J. All of these factors ensure that Tapestry will persist in its evolution and will
continue to improve the ways in which it makes the life of web developers easier
and more productive.
This book will allow you to delve into a different, better world of web development than the one you have known. It will show you an innovative approach for
creating and organizing your applications, enable you to develop more robust
and scalable code, and make previously difficult tasks much simpler. After you
get to know Tapestry, you will start looking at web development in a very different way. Enjoy!
THEODORE TONCHEV
“MINDBRIDGE”
TECHNICAL DIRECTOR, RUSHMORE DIGITAL
preface
From the fall of 1998 through late 1999, I worked on a mix of interesting
projects for my then-employer, Primix, a Boston Internet consulting firm. The
first of these was a standard CRUD (Create Read Update Delete) application,
and it used servlets and JavaServer Pages (JSPs). Despite the fact that we were
a small team (just three or four of us at various times) working on a very modest application, it seemed like we were constantly having to reinvent the wheel.
I’d had some exposure to Apple’s WebObjects framework, and the difference
between the two approaches was like night and day: Not only did WebObjects
have better tool integration, including a WYSIWYG (What You See Is What You
Get) HTML editor, it was easier to use even without the extra tools, just by editing the underlying template and configuration files. We were a small, smart,
experienced team—but working with servlets and JSPs was always an uphill
battle compared to using WebObjects.
A major part of this problem was impedance; each developer worked in his
or her own corner of the application and had a particular way of working. Different developers used different naming conventions and solved similar problems in different ways—which caused extra grief when new developers cycled
through the team, or when we had to venture outside our little fiefdoms to fix
bugs. Remember, this was a time before J2EE servers really existed, before custom JSP tag libraries, before Sun had even started promoting the Model 2 design
pattern. It was also obvious that WebObjects, with its notion of components and
xix
xx
PREFACE
all the magic it did to hide URLs from the developer, was a far superior option.
At the same time, steep licensing fees and lukewarm support and marketing by
Apple made pitching a WebObjects solution to clients a nonstarter. We soldiered
on with servlets and JSPs.
Near the end of 1999 I was left with engagements in the pipeline but no
short-term responsibilities. What was to eventually become Tapestry started as a
kind of mental doodle: How does WebObjects do it? I started drawing diagrams
and sketches, puzzling out a way to gracefully connect stateless HTTP requests to
stateful Java objects via Java servlets, and eventually decided to code up a little
prototype to see if my ideas would work. Somewhere along the line, I came up
with the name: Tapestry.
Tapestry quickly took on a life of its own; as primitive as that early code base
looks compared to the Tapestry of today, it was still a solid foundation, something
that would start to address the issues Primix had with servlet and JSP development. Primix was in the business of creating high-quality web sites on a regular
basis, and Tapestry was a platform that would support the solid software engineering principles necessary to achieve its goals. Tapestry set up shop at SourceForge (http://sf.net) and was released under the Lesser GNU Public License.
Along the way, I realized something profound: I could make a difference. Interesting software didn’t have to come forth directly from the Suns, IBMs, and
Microsofts, or even from the Apaches; it could be homegrown, like Tapestry.
With Tapestry, I really thought I could change the world…or at least, the little
niche of the world that writes web applications in Java.
As fun as working on the code has been, the most fascinating part of Tapestry
has been the creation of a real, online community around Tapestry. These are
people for the most part I have yet to meet, who have recognized Tapestry as a
Good Thing and who work as hard as I do to improve and promote it. It was on
the strength of this community that Tapestry relocated from SourceForge to
Apache’s Jakarta project, alongside established projects such as Struts, Turbine,
and Velocity.
Crafting a book about Tapestry has been more difficult and more intense an
experience than anything else related to the project, but ultimately the result is
worth the effort. What you are holding in your hands may open up a whole new
approach to web application development for you and (I hope) will change your
world for the better.
acknowledgments
Working on this book has been the highlight of my career to date. I’ve never
before had the opportunity to be so focused and so self-directed for such a long
time. I would never have had this opportunity without the support and friendship of Timothy Dion, the CTO of Primix, who gave me the green light to work
on Tapestry in the first place, as well as the support to release it as open source.
Just as instrumental was my good friend Gregory Burd, who championed Tapestry within Primix and put it into play for its first big engagement.
The Tapestry crew—Theodore Tonchev, Richard Lewis-Shell, David Solis,
Neil Swinton, Harish Krishnaswamy, and Geoff Longman—have been tireless
supporters of Tapestry, and good friends, many of whom I’ve yet to meet in
person! Thanks as well to Andrew C. Oliver for sponsoring Tapestry in the
Apache Jakarta project.
Writing a book is a group effort—without the tireless help from the crew at
Manning, this book would never have reached your hands. Both Marilyn
Smith and Liz Welch provided endless assistance with grammar and style, as
well as much valuable advice. Tiffany Taylor’s eagle eyes spotted an embarrassing number of typos and inconsistencies. Denis Dalinnik did a standout
job of typesetting. Special thanks to Jackie Carter, for keeping me honest and
keeping me focused on the reader. Thanks also to my technical reviewers, especially Bill Lear, Ryan Cox, Adam Greene, and Joel Trunick—they pushed me
to create a better book than I would have thought possible. Thanks as well to
xxi
xxii
ACKNOWLEDGMENTS
Mary Piergies for keeping it all organized, to Clay Andres for getting the whole
thing started, and to publisher Marjan Bace for the vote of confidence.
Special thanks go to my wife, Suzanne, who has encouraged me every step of
the way on this long process. Suzanne’s support has never wavered, even as this
project consumed my time, attention, energy, and the dining room table (my
workspace) for months on end. Suzanne has indulged my need to talk about
Tapestry and this book long past the point where anyone else would have run
screaming from the room, and I love her for that, and for her boundless enthusiasm. We are stylin’, baby!
about this book
Tapestry is a comprehensive web application framework for the Java programming language. Tapestry is based on components, highly reusable building blocks that can be quickly and easily combined to form pages within your
application. By using and reusing components, and creating your own components, you can create richly interactive, robust applications with only a
modest effort.
Tapestry’s basic style is to break problems into smaller and smaller units;
this complements a team development environment where different developers work on different parts of the application. The framework makes it easy
for both Java and HTML -only developers to work together without accidentally undermining each other’s work.
When building a web application with any technology, you will be faced
with a constant stream of questions: How do I figure out what the user has
requested? Where can I store this bit of information? How can I safely add this
new functionality? How can I make my application scale? In too many environments, it’s easy to make the wrong decision when confronted with any of
these, and many other, development-time questions. It’s too easy to take a
quick-and-dirty detour down the wrong path, which ultimately comes back to
bite you when you are least prepared to deal with it.
The central goal of Tapestry is to make the easiest choice the correct choice.
Over the course of this book, we’ll show you how to build applications using
xxiii
xxiv
ABOUT THIS BOOK
Tapestry, but we will also show you the hidden traps and tangles that Tapestry
helps you to avoid.
Roadmap
■
■
■
■
■
■
■
Chapter 1 introduces web applications in general. We begin to describe
where Tapestry fits into the overall scheme of things and expand on the
basic goals achieved by the framework.
Chapter 2 sets the stage, describing the implementation of a simple web
application that plays the word game Hangman. Here we introduce many
of the major concepts of Tapestry.
Chapters 3 through 5 describe how Tapestry handles HTML form input,
including the framework’s sophisticated form-validation subsystem.
Chapter 6 describes how to build basic reusable components and how to
package them into component libraries.
Chapter 7 delves into the inner implementation of Tapestry, shedding
light on how the framework addresses scalability issues, localization, and
the lifecycle of pages.
Chapter 8 describes advanced components, including components that
create client-side JavaScript. We also discuss integrating Tapestry with a
traditional servlets-and-JSPs application.
Chapters 9 and 10 describe the Virtual Library, a complete J2EE example
application using Tapestry for the presentation layer and a mix of session
and entity EJBs for the application layer. We include several real-world
examples that illustrate how to build and polish a Tapestry application.
Who should read this book?
This is a book about getting things done using the Tapestry framework; as such,
it will appeal to Java web developers looking for a better, easier way to build web
applications. Because Tapestry is explicitly designed to support team development of web applications, this book will also be of interest to managers looking
for a better way to leverage their team’s efforts.
This book is targeted at people who have at least gotten their feet wet in terms of
Java web application development (or perhaps have already taken the full plunge).
Therefore, we assume that you are at least somewhat acquainted with a number of
concepts and technologies. Obviously, an understanding of the Java programming
language is a prerequisite, as well as familiarity with such key Java APIs as the
ABOUT THIS BOOK
xxv
collections framework. You should also be clear on the distinction between Java
interfaces and Java classes.
Much of Tapestry concerns the moving of information from one object to
another; this is facilitated using JavaBeans properties. The core concept of the
JavaBeans framework is that an object can be treated as if it was a Map, as a collection of named properties that can be read or updated without knowing the actual
class of the object. More information about JavaBeans is available at http://
java.sun.com/products/javabeans/docs/.
You should be familiar with the basic set of HTML tags, including <body>,
<img>, <a>, <form>, and <input>. You must also be familiar with URLs and query
parameters and the difference between HTTP GET and HTTP POST.
Many of the artifacts of a Tapestry application are XML documents. You
should be familiar with basic XML usage and syntax.
Some of the later examples show how to implement client-side logic. This
requires an understanding of JavaScript (the scripting language that executes
within a client web browser) as well as the Document Object Model, the data
structure representing a web page inside a web browser.
Code conventions and downloads
This book includes copious examples, which include all the Tapestry application
artifacts: Java code, HTML templates, and XML specification files. Source code in
listings or in text is in a fixed width font to separate it from ordinary text. Additionally, Java method names, component parameters, object properties, and
HTML and XML elements and attributes in text are also presented using fixed
width font. Java method names will generally not include the signature (the list
of parameter types).
Java, HTML, and XML can all be quite verbose. In many cases, the original
source code (available online) has been reformatted, adding line breaks and
reworking indentation, to accommodate the available page space in the book. In
rare cases, even this was not enough, and listings will include line continuation
markers. Additionally, comments in the source code have been removed from
the listings.
Code annotations accompany many of the source code listings, highlighting
important concepts. In some cases, numbered bullets link to explanations that
follow the listing.
Tapestry is an open-source project, released under the very liberal Apache
Software License. Directions for downloading Tapestry, in source or binary form,
xxvi
ABOUT THIS BOOK
are available from the Tapestry home page: http://jakarta.apache.org/tapestry/.
Documentation available from the home page also identifies how to download
the source code via CVS so that you can build the framework locally, if you are
so inclined.
The Tapestry distribution includes the Virtual Library application described
in chapters 9 and 10.
The source code for all examples in this book is available from Manning’s web
site: www.manning.com/lewisship/. To run the examples, you need to download
Tapestry and the Tomcat servlet container. Appendix B contains the details.
Author Online
The purchase of Tapestry in Action includes free access to a private web forum run
by Manning Publications, where you can make comments about the book, ask
technical questions, and receive help from the author and from other users. To
access the forum and subscribe to it, point your web browser to www.manning.com/lewisship. This page provides information on how to get on the forum
once you are registered, what kind of help is available, and the rules of conduct
on the forum.
Manning’s commitment to our readers is to provide a venue where a meaningful dialogue between individual readers and between readers and the author
can take place. It is not a commitment to any specific amount of participation
on the part of the author, whose contribution to the forum remains voluntary
(and unpaid). We suggest you try asking the author some challenging questions,
lest his interest stray! The Author Online forum and the archives of previous
discussions will be accessible from the publisher’s web site as long as the book is
in print.
About the author
Howard Lewis Ship is the lead developer for the Tapestry project. He cut his
teeth writing customer support software for Stratus Computer, but eventually
traded PL/1 for Objective-C and NextStep before settling into Java. Howard is
currently an independent open-source and J2EE consultant, specializing in customized Tapestry training. You can find Howard on the web at http://howardlewisship.com. In the real world, he lives in Quincy, Massachusetts, with his wife
Suzanne, a novelist.
about the title
By combining introductions, overviews, and how-to examples, the In Action
books are designed to help learning and remembering. According to research
in cognitive science, the things people remember are things they discover during self-motivated exploration.
Although no-one at Manning is a cognitive scientist, we are convinced that
for learning to become permanent it must pass through stages of exploration,
play, and, interestingly, re-telling of what is being learned. People understand
and remember new things, which is to say they master them, only after actively
exploring them. Humans learn in action. An essential part of an In Action guide
is that it is example-driven. It encourages the reader to try things out, to play
with new code, and explore new ideas.
There is another, more mundane, reason for the title of this book: our
readers are busy. They use books to do a job or solve a problem. They need
books that allow them to jump in and jump out easily and learn just what they
want just when they want it. They need books that aid them in action. The
books in this series are designed for such readers.
xxvii
about the cover illustration
The figure on the cover of Tapestry in Action is a “Gonaquesa Baylando,” a
dancing woman of the Gonaqua tribe in Africa. The Gonaquas were herders
and farmers living on the southern coast of the continent near the Cape of
Good Hope in what is today South Africa. The illustration is taken from a
Spanish compendium of regional dress customs first published in Madrid in
1799. The book’s title page states:
Coleccion general de los Trages que usan actualmente todas las Nacionas del Mundo
desubierto, dibujados y grabados con la mayor exactitud por R.M.V.A.R. Obra muy
util y en special para los que tienen la del viajero universal
which we translate, as literally as possible, thus:
General collection of costumes currently used in the nations of the known world,
designed and printed with great exactitude by R.M.V.A.R. This work is very useful
especially for those who hold themselves to be universal travelers
Although nothing is known of the designers, engravers, and workers who
colored this illustration by hand, the “exactitude” of their execution is evident
in this drawing. The “Gonaquesa Baylando” is just one of many figures in
this colorful collection. Their diversity speaks vividly of the uniqueness and
xxviii
ABOUT THE COVER ILLUSTRATION
xxix
individuality of the world’s towns and regions just 200 years ago. This was a time
when the dress codes of two regions separated by a few dozen miles identified
people uniquely as belonging to one or the other. The collection brings to life a
sense of isolation and distance of that period—and of every other historic period
except our own hyperkinetic present.
Dress codes have changed since then and the diversity by region, so rich at
the time, has faded away. It is now often hard to tell the inhabitant of one continent from another. Perhaps, trying to view it optimistically, we have traded a cultural and visual diversity for a more varied personal life. Or a more varied and
interesting intellectual and technical life.
We at Manning celebrate the inventiveness, the initiative, and, yes, the fun of
the computer business with book covers based on the rich diversity of regional
life of two centuries ago‚ brought back to life by the pictures from this collection.
Part 1
Using basic Tapestry
components
C
hapters 1 through 5 introduce you to the basic concepts of Tapestry.
You’ll learn how Tapestry defines an application, a page within an application,
and a component within a page. These chapters gradually expose you to the
more complicated challenges in developing web applications and explain how
Tapestry meets these challenges.
This chapter covers
■
Understanding web applications
■
The goals of Tapestry
■
Using the Model-View-Controller pattern
■
Essential Tapestry classes and interfaces
3
4
CHAPTER 1
“… and we’re going to implement it as a web interface,” your boss announces, and
you start to get that sick feeling in your stomach. You had only been half-listening,
daydreaming about the clever, graceful user interface you were going to create. But
now you’re in unknown territory; you might as well be fresh out of school! Your mental roadmap to the application starts to lose focus; you can still see the islands of code
and functionality, but now you imagine a dark, churning sea between them. How
are you going to get this to work as a web application?
If you are used to creating traditional desktop applications, switching over to developing web applications can be a journey into unfamiliar territory. Web applications
turn everything you know upside down—in desktop applications, your code is in
the driver’s seat, ultimately controlling every pixel visible to your single user, and
the application is all-knowing about every mouse movement or key press. In web
applications, your code is just part of an overall picture involving a variety of computers and network communications… and your code must support multiple simultaneous users. This book describes how to use the Tapestry web application
framework to ease your transition into the web application development space.
The Tapestry framework is more than a crutch for new web developers; even
seasoned developers and project leaders will find advantages to building their
applications using it. Tapestry allows true black-box code reuse within and
between applications; it addresses the concerns of large teams of developers with
mixed skill sets; it has support for application internationalization as a central
feature, not an afterthought. This book shows you how to leverage Tapestry to
produce a more sophisticated, more robust application faster and more easily
than you might suspect.
Tapestry is an open source, Java web application framework designed from
the ground up to help you deliver the best web applications possible. It allows
new developers to get up to speed easily, creating working applications with a
surprisingly small amount of Java code. It includes numerous features that
address the issues encountered during team development of large, complex,
internationalized applications. Tapestry is open-ended—it doesn’t shoehorn you
into any one category of web application, which means you can build your own
Amazon, Slashdot, Yahoo!, or eBay with it. Tapestry is designed to plug into
whatever form of back-end systems you use, with a minimum of fuss.
This is a framework that lets you have it both ways: you create a web application, but you don’t have to throw away everything you’ve previously learned
about coding just because you are creating a web application instead of a traditional desktop application. The cornerstone of Tapestry is letting you write
What are web applications?
5
applications in terms of objects, methods, and properties—not in terms of URLs
and query parameters and Java servlets.
1.1 What are web applications?
In the earliest days of the World Wide Web, the majority of web sites were
entirely static, unchanging. Many early sites were created entirely by hand, by
folks using simple text editors to directly edit the HTML of individual pages.
Other sites were created in a batch mode, where a source file or database would
be transformed into HTML. In fact, the original purpose of the Web was to
allow physicists to easily share their publications; physics papers written using
LaTeX (a special-purpose typesetting language) were translated into HTML for
instant sharing.
Web sites didn’t stay static for long; instead, they transformed into web applications. A web application is interactive; the end user will see links that may be
clicked and HTML forms that may be filled out. These links and forms become
requests sent to a server, which will respond with a new HTML page, often created (from a template) on the fly, as a personalized response to a specific request.
Figure 1.1 illustrates the general flow of a web application, which is divided
into four steps:
1
A request is received by the server. Information in the request is used to
dispatch control to the correct application code.
2
Application-specific code executes. This code interprets the information
available to it in the URL and query parameters, and executes some
application-specific logic. This includes deciding what response to send
back to the client.
Figure 1.1
Web applications are built
around a cycle; a request is
received, applicationspecific code is executed,
and a response is rendered,
eventually appearing in the
client’s web browser.
6
CHAPTER 1
3
A response is rendered. The response includes links and forms that will
result in new request cycles.
4
The response is visible in the client web browser. Users now have a
chance to see the results of the action they initiated. In addition, any
client-side logic is activated.1 The user may click links or submit forms,
resulting in a new request cycle.
Web applications are written in the context of the underlying network protocol
connecting the client web browser (such as Microsoft Internet Explorer or
Netscape Navigator) and the web server: the Hypertext Transport Protocol
(HTTP). HTTP is a stateless protocol, meaning that each request is completely
independent of any requests that come before or after it. There is no explicit
concept of user identity in HTTP or even, at the protocol level, any way to identify
that a series of requests comes from a single user. The stateless nature of HTTP is
an important factor in its success; a stateless protocol is much easier to implement in a web browser or a web server.
For dynamic, interactive web applications, the inherent statelessness of HTTP
creates new challenges. End users are not concerned with this; they simply
expect to progress from a catalog page to a shopping cart page to a checkout
page. Creating a stateful application from a stateless protocol is a concern for the
application developer and involves issues not present in a desktop application.
By comparison, imagine writing a document by launching a text editor application, loading a file, typing a single sentence, saving the file, and then shutting
down the application before repeating the whole process for the next sentence.
For a web application, each request is like one sentence in that document…and
part of the request had better be which document is being edited and where in the
document the new sentence goes. A desktop application automatically has a kind
of continuity that must be specifically engineered into a web application.
This same basic web application flow can be implemented in many different
environments and languages. In the early heyday of web applications, scripting
languages such as Perl were most often used to implement the applicationspecific code. Over time, as web applications evolved from clever novelties into
critical enterprise infrastructure, the implementation choice for web applications
shifted to more powerful, higher-level languages. The Java programming
1
This takes the form of JavaScript programs embedded within the HTML that execute within the client’s web browser.
7
language is especially well suited for web application development because of its
standard web application framework, the Java Servlet API.
1.2 What are Java servlets?
The Java Servlet API is an open standard, created by Sun, for creating web
applications using the Java programming language. A servlet is an object
responsible for receiving a request from a client web browser and returning a
response—an HTML page to be displayed in the browser. The Servlet API
defines an interface and base class for servlets, as well as interfaces for several
other supporting objects, such as the HttpServletRequest (which represents a
request and allows the servlet to access query parameters). Vendors, both open
source and proprietary, provide the actual server code and the implementations
of the standard interfaces.
Servlets are most often paired with JavaServer Pages (JSPs) in order to generate a response. JSPs are a standard templating technology for servlets. A JSP is a
mix of ordinary, static HTML with additional, specialized tags and directives that
provide dynamic output, such as including the current user’s name or the contents of an online shopping cart. Under the hood, each JSP is converted into a
Java servlet that is compiled and loaded into memory. More information about
servlets and JSPs is available online and in print, including Web Development with
JavaServer Pages.2
A servlet operates within a servlet container. The servlet container serves as a
bridge between HTTP and the Java servlet code that you, as the developer, will
write. The servlet container is responsible for instantiating and initializing your
servlets as needed. Servlet containers may be standalone, such as Apache Tomcat, or may be one part of an overall application server, such as BEA WebLogic,
IBM WebSphere, or the open source JBoss application server.
The servlet container is responsible for selecting the correct servlet to invoke
based on the URL of the request; a single web application will contain many servlets. The web application’s deployment descriptor (an XML file packaged with the
application) gives the name and Java class for each servlet and identifies which
URLs are mapped to which servlets. For example, your application may have a registration page containing an HTML form enabling new users to register (supplying their name and email address or other data). That form will submit to the
2
Duane K. Fields, Mark A. Kolb, and Shawn Bayern, Web Development with JavaServer Pages, 2nd Edition
(Greenwich, CT: Manning Publications, 2001).
8
CHAPTER 1
Figure 1.2 Incoming requests to the server are dispatched to a particular servlet instance. The
servlet forwards to a JSP that renders the response sent back to the client web browser.
/addCustomer URL, which is mapped to the addCustomer servlet, which is instantiated as the AddCustomerServlet class. The deployment descriptor will include
the following elements describing the addCustomer servlet’s configuration:
<servlet>
<servlet-name>addCustomer</servlet-name>
<servlet-class>com.mycompany.AddCustomerServlet</servlet-class>
</servlet>
<servlet-mapping>
<servlet-name>addCustomer</servlet-name>
<url-pattern>/addCustomer</url-pattern>
</servlet-mapping>
Figure 1.2 shows how the application request cycle applies to servlets.
As can be seen from this example, each servlet represents a particular operation within the overall web application. The AddCustomerServlet class will include
code that reads the query parameters submitted with the form and uses those
values to create some kind of Customer object that can be stored into a database.
Additional logic must decide what form of response to display—normally a confirmation page, but possibly the original page containing the form if an error
occurred when creating the customer.
This operation-centric approach can work well for many parts of a web application. In this example, there is only one registration page, and so only one
place from which the addCustomer servlet can be invoked. If the original registration page must be redisplayed (to present the user with an error message), there
is no guesswork about what page to use; it will always be Registration.jsp.
When an operation can be associated with many different application pages,
the operation-centric approach can start to become a burden. For example,
many applications include pages that display long lists of information and allow
9
users to navigate through the results one page at a time. Such pages include
links to operations such as “next page of results” or “previous page of results.”
Using servlets, it becomes necessary to somehow identify not just the operation, but to where the operation should be applied. Are we displaying the product catalog page, or the shopping cart page, or the related-items page? Now we
need a way to identify the page; a common approach is to add a query parameter to the URL to identify that page.
Multiplexing operations onto multiple pages is just one factor a servlet developer must keep in mind. Another is dealing with the multithreaded environment
in which a servlet object operates.
1.2.1 Understanding servlet multithreading
Within a servlet container each servlet is instantiated exactly once; all requests
mapped to the servlet’s URL pattern will be passed through to the service()
method of a single servlet object.3
Servlet containers are multithreaded, which is to say they contain multiple
threads of execution, allowing requests from several client web browsers to be
processed simultaneously—a critical factor for building scalable web applications. Despite this, servlets are still single instances, shared between the threads.
Servlet instances must be thread-safe; instance variables can’t be used either for
temporary storage during a single request or for persistent storage between
requests, because any information stored in an instance variable is likely to be
overwritten by other threads.
Servlets are the odd-man-out in the object-oriented world; they are technically still objects but largely fall short of the accepted definition of an object: a
combination of operations and state. Servlets aren’t allowed to directly store state
in their own instance variables. Local variables are fine for some things, but
there are specific cases where local variables are not applicable, such as forwarding to a JSP to render a response.
JSPs, like servlets, are stateless and multithreaded; therefore, there’s no way
that a servlet could, for example, set a property of a JSP to the current user’s
name. Instead, the Servlet API object HttpServletRequest is used to store such
transient information. One aspect of the request object is that it can be used to
store named attribute values. The servlet can store values into the request before
forwarding to a particular JSP to render the response, as shown in figure 1.3.
3
Your servlet code will be inside the doGet() or doPost() method that is invoked by the service() method.
10
CHAPTER 1
Figure 1.3 A servlet can store a request attribute before forwarding to the JSP; the
JSP can then read the attribute and incorporate it into the response.
This is an example of the servlet “pushing” data into the JSP via the HttpServletRequest. The servlet and JSP are intimately tied together; they must agree on
exactly which attributes will be stored into the request object, and there’s no
room for error on the naming (username is not the same as userName, for example). Overall, the HttpServletRequest provides a flexible solution for managing
transient information; as we’ll see, a similar solution addresses the more complicated problem of persistent server-side state.
1.2.2 Managing server-side state
The single attribute that most widely separates a static web site from a true web
application is server-side state. Amazon has a shopping cart that remembers what
items you’ve placed in it from one request to the next. Google remembers enough
about your query to allow you to page through the list of results. Any application
will need to identify you as a particular user and remember some amount of
information about the activities you’ve performed while visiting the site.
For a web application to be truly interactive and personalized, it must have a
way of identifying the user from one request to request, and a way to store information on the server about that particular user. This is not necessarily the same
thing as authentication; the application may not be able to match the requests to
11
a particular user account or credit card number, but that shouldn’t prevent the
application from tracking short-term information, such as the contents of a
shopping cart in an e-commerce application—information that needs to persist
from one request to the next.
So, if HTTP is stateless, how can an application store any state at all? What’s
needed is some form of identifier to be created on the server, sent back to the client, and returned to the server as part of subsequent requests. Fortunately, most
client web browsers support HTTP cookies; a cookie is a small string stored within
the browser. The server can invisibly include a cookie value in a response,4 and
the client browser will send the cookie value back up to the server in each subsequent request—exactly what is required to support server-side state. A series of
such linked requests forms a session.
For servlets, server-side state is stored in the HttpSession object. Like the
HttpServletRequest, the session object may store named attribute values; the difference is that such values are stored on the server between requests. For example,
the ShoppingCart object created and stored in the session by the addToShoppingCart servlet will later be available to the checkout servlet responsible for collecting payment information, even though the user may do quite a bit of browsing
around the product catalog in between those two operations.
Sessions are not created until specifically requested; the HttpServletRequest
object includes a method, getSession(), for creating the session as needed.
Some people or organizations do not allow HTTP cookies (most browsers allow
support for cookies to be turned off, for privacy and security reasons). With a little extra effort, it is still possible to support stateful servlet applications even when
the client does not support cookies.5
Support for servlets and JSPs is a standard feature for all popular application
servers, and many developers do manage to create successful web applications
using just the standard servlet technology. At the same time, there is a growing
acknowledgement that servlets are just the starting point for Java web applications, as evidenced by an increasing number of popular frameworks. This goes
beyond just Tapestry and includes other frameworks such as Turbine (http://
jakarta.apache.org/turbine/), Maverick (http://mav.sourceforge.net/), WebWork (http://
www.opensymphony.com/webwork/) and, most prominently, Struts.
4
5
The HTTP cookie is part of the HTTP “envelope” around the HTML “message”; it isn’t part of the
HTML and won’t be visible even if the user views the HTML source of a rendered page.
The servlet container can encode the session ID into the application’s URLs to provide session continuity.
12
CHAPTER 1
1.2.3 Using Struts with servlets
Struts, like Tapestry, is an open source framework available from the Jakarta
project, at http://jakarta.apache.org/struts/. Struts is an extension to standard
servlets, rounding out rough edges and adding a few much-needed abstractions
that simplify servlet coding. A comprehensive guide to Struts is available in Struts
in Action.6
A Struts application uses a single servlet to represent all possible operations.
The servlet uses a set of Struts-specific configuration files to dispatch incoming
requests to a particular Struts Action object. Like servlets, Struts Actions are multithreaded and stateless. Struts also includes a set of JSP tag libraries that simplify and standardize the creation of dynamic web pages, especially with respect
to creating URLs for links and form submissions.
Struts uses its configuration files to loosely couple Actions to the JSPs that
render responses (and makes it easier to use another templating system, such as
Velocity). JSPs are assigned logical names in the Struts configuration that can be
referenced from an Action; if the JSP is moved or renamed, the side effects are
isolated to just the Struts configuration file, not the Java code for the Action.
The same mechanism also makes it easier to chain together a series of Actions;
for example, the one Action can be used to update some server-side state and
can forward to a second Action that stores information into the HttpServletRequest before forwarding to a JSP to render the response; this useful pattern is
easier to accomplish in Struts than with ordinary servlets.
Struts includes a standard mechanism for converting form submissions into
objects, which works by matching query parameter names against JavaBean
properties of a user-defined form object. The latest release of Struts complements this with a rules-based input-validation system.
Struts standardizes some common patterns and techniques that would otherwise need to be re-invented by each developer, but it is still fundamentally similar
to ordinary servlet development. Whether developing using just servlets or with
servlets augmented by Struts, you will still encounter some significant limitations
in the basic development model that you must overcome.
1.2.4 Understanding the limitations of servlets
There is a qualitative difference between the kind of servlet application seen in
demos and tutorials and true enterprise applications. Demo applications are
6
Ted Husted et al., Struts in Action (Greenwich, CT: Manning Publications, 2002).
13
small, focused, and limited in scope; they are often created in a short period of
time by a single individual, who fills the roles of application architect, HTML
developer, and Java developer simultaneously.
By contrast, development of a full-scale application entails a number of realworld concerns:
■
A large number of developers (HTML and Java) may be working in parallel.
■
Individual developers will have varying levels of skill and understanding.
■
The HTML portions of the application may be developed by team members with little or no knowledge of Java or JSPs (or even by a completely
separate, isolated team).
■
Large applications (with hundreds of distinct pages) may be so complex
that individual developers will understand only a small portion of the
overall application.
■
Successful applications will, almost by definition, grow and expand in complexity to meet new customer requirements.
These concerns manifest themselves in a number of common antipatterns, which
we describe next.
Weak binding
The most fundamental issues with building, maintaining, and extending a web
application using servlets surround two related problems: weak binding and
unwanted dependencies. In a servlet application, the connections between pages
are weak because those connections are expressed as URLs, not as method invocations or object properties. At the time that a JSP renders a link (or form) connecting to another page in an application, all it is doing is outputting a string as
the href attribute of an HTML <a> hyperlink tag (or the action attribute of an
HTML <form> tag). This is another difference between desktop applications and
servlet applications: Changing a method signature in an object of a desktop
application will create errors in code elsewhere in the application that uses the
old method signature. The integrated development environment ( IDE ) will
clearly show those errors, and you can find and fix all of them before running
and testing your application.
In a servlet application, the connection between two servlets is reduced to a
string within a JSP. The JSP must include a URL string (for the target servlet) in
the response sent to the client web browser. Where does that URL string come
from? The developer manually inserts it into the JSP, after consulting the web
14
CHAPTER 1
deployment descriptor (or Struts configuration file) to find the operation mapping to link to. Assuming the developer introduces no typos, the link between
pages should be functional—but there’s no way to be absolutely sure without actually running the application. This servlet-to-servlet linkage is a weak binding;
the Java compiler or other tools cannot validate (at build time) that the URLs are
correct and functional; validation can occur only at runtime.
Weak URL bindings don’t really show their teeth until the application is
altered in some way, such as a change to the signature of an application operation; for example, an operation URL may change (i.e., /addCustomer to /addRetailCustomer), or the query parameters passed in the URL may change (i.e., productId
to sku). Such a change in signature will require a hunt throughout the application
for references to the old operation signature, to update it to the new signature.
In other words, there are unwanted dependencies between the pages that link to
operations and the implementations of the operations themselves.
Team conflicts
Weak bindings can provoke one form of team conflict; another example occurs
between the HTML developers and the Java developers. It is too easy for an
HTML developer to make a minor change to a JSP that “breaks” the page in
some way. In most environments, the entire application would need to be built,
deployed, and executed in order for the offending change to be noticed. Only
recently have JSP-aware HTML editors become available; when used, these editors can provide a real-time WYSIWYG view of application pages—but it is not a
given that HTML developers will have access to such tools or be fully proficient
using them. This can leave HTML developers in a situation where they need to
build, deploy, and run the live application to see the effects of their changes—a
solution that may not be practical unless the HTML developers are co-located
with the Java developers.
Another form of team conflict concerns choke points in the application—shared
files or Java classes that many Java developers will need to frequently update. A
key example of this is the web deployment descriptor, the file used to define servlets and URL mappings. As each developer adds a new servlet, the file must be
updated; with many developers doing similar work in parallel, the opportunity
for conflicts is inevitable. Struts reduces the level of contention for the web
deployment descriptor by allowing multiple configuration files to express the
same kind of information, but there’s still room for conflicts elsewhere.
15
Bad coding shortcuts
Creating servlet applications can be a code-intensive process; this can lead to
developers taking bad coding shortcuts. A common example is merging database access code directly into a JSP. Initially this seems like a good idea, since it
locates the code for database access directly with the JSP that will present the
results of the query to the user.
Unfortunately, this approach leads to horrific problems in a running application. It makes the JSP very fragile in the face of any change to the underlying
data model. The query will simply throw an exception at runtime if a table or
column name changes. It is unlikely that shortcut code will take advantage of
database connection pools, but will instead create, use, and close a connection
inline. While this works fine in testing, it falls flat in the face of a high volume
of concurrent users. Finally, a shortcut approach will likely not include a
try…finally block to ensure that the connection is closed when done; this can
lead to database connection leaks, another way to bring a production application
to its knees.
Here are other common coding shortcuts:
■
Failure to properly localize output in an internationalized application
(including failure to use localized date and number formatters).
■
Not using HttpServletRequest.encodeURL() to encode the session ID into
application URLs, which breaks the application for users who do not have
HTTP cookies enabled.
■
Not properly filtering output to convert reserved HTML characters (such
as < and >) to HTML entities; this results in visual oddities in the client
web browser.
■
Using short, unqualified names for HttpServletRequest and HttpSession
attributes, resulting in naming conflicts and overwritten data.
Many other examples abound but require considerably more background material to fully describe. The point of this section was not to launch a diatribe
against servlets—servlets are a powerful tool for creating web applications.
Instead, we want to focus on the areas where servlets require too much effort or
expertise on the part of the developer to be used effectively, especially in large
and complex applications.
16
CHAPTER 1
1.3 Why do we need Tapestry?
Tapestry exists to simplify Java web application development. All the features of
this framework are designed to make it simpler to create robust applications that
are easier to construct, debug, maintain, and extend than traditional servlet
applications. Tapestry extends your reach, allowing you to create more powerful,
more useful, more interactive applications faster than you would using ordinary
servlets. In addition, you’ll have greater confidence that your application will be
free of bugs. Also, Tapestry reduces the level of contention between team members (both Java and HTML) working on a shared project.
Ultimately, Tapestry applications are still servlet applications. With enough
code, and enough time to debug that code, there’s nothing (from an end user’s
point of view) that a Tapestry application can do that a servlet application can’t.
Tapestry applications still follow the request cycle in figure 1.1; a request is
received, some kind of server-side processing occurs, and a response is sent back
to the client web browser. The difference is that Tapestry allows you to write far
less code, and the code you do write is simpler and more natural because Tapestry excuses you from concerns about multithreading.
Tapestry recasts the stateless, operation-centric Servlet API into a stateful,
component-centric model, the familiar coding pattern used in traditional desktop applications. The approaches you are accustomed to taking—such as combining operations and data together to form objects, using components and
inheritance, storing information in instance variables, and so forth—all of these
are hallmarks of Tapestry application development as well. This will not seem
remarkable if you haven’t done any work with servlets before; if you have, you’ll
be surprised by the things you don’t see in a Tapestry application: the HttpSession
and HttpServletRequest objects, or any thought of URLs or query parameters;
all of these things are pushed down into the bowels of the Tapestry framework,
where the details are handled robustly, efficiently—and invisibly.
1.3.1 What is a framework?
A framework is a set of cooperating classes that make up a reusable design for a
specific category of software. A framework is different from a toolkit; a toolkit is a
collection of individually reusable classes, each performing a small, isolated
function that is applicable to a wide variety of uses. By contrast, a framework
…dictates the architecture of your application. It will define the overall structure,
its partitioning into classes and objects, the key responsibilities thereof, how the
classes and objects collaborate, and the thread of control. A framework predefines
17
these design parameters so that you, the application designer/implementer,
can concentrate on your application. The framework captures the design decisions that are common to its application domain.7
Frameworks are very useful; instead of your having to start with a clean slate, the
design is partially filled in and the path to follow is clear. Many design decisions
are already made for you, decisions that leverage the combined experience of
the frameworks’ authors and users. A framework also provides a significant coding head start in the form of reusable classes, utility classes, and base classes for
you to extend. Ideally, you simply have to fill in the blanks, connect a few dots,
and provide some application-specific subclasses. The end result is less code for
you to write and more consistency between applications, not just from the developer’s point of view but from the end users’ perspective as well.
Many frameworks, including Tapestry, incorporate a component object
model, which we will look at in the next section. This approach allows you to easily mix and match off-the-shelf objects with objects you create yourself.
1.3.2 What is a component?
A component is an object that fits into an overall framework; the responsibilities of
the component are defined by the design and structure of the framework. A
component is a component, and not simply an object, when it follows the rules of
the framework. These rules can take the form of classes to inherit from, naming
conventions (for classes or methods) to follow, or interfaces to implement. Components can be used within the context of the framework. The framework will act as
a container for the component, controlling when the component is instantiated
and initialized, and dictating when the methods of the component are invoked.
A component framework is a useful thing; components, more so than objects, can
be easily combined to perform complex operations or to create entirely new components. The rules of the framework dictate how components can be used together. A
component object model is the portion of the framework that defines the rules for how
individual components may be combined for such purposes. The framework’s component object model also dictates the responsibilities and lifecycles of containers
(components that are constructed from other components) and contained objects.
The JavaBeans framework is a good example of a component framework—
it’s very general and adaptable. Following the basic naming convention for
7
Erich Gamma et al., Design Patterns: Elements of Reusable Object-Oriented Software (Reading, MA: AddisonWesley, 1994).
18
CHAPTER 1
methods specified by the framework (the well-known getters and setters naming
conventions) makes it possible for components created at different times and by
different authors to be seamlessly combined together at runtime. The JavaBeans
framework’s component object model is structured around an event-notification
system. Individual components may be event producers or event consumers.
Another standard naming convention allows objects and components to register
for event notifications in a generic fashion.
Originally, the JavaBeans framework was intended for use in developing
graphical user interfaces (GUIs) for desktop applications; the components were
expected to be visual objects, such as buttons and text fields, that could be
manipulated in a visual GUI editor. Ultimately, the JavaBeans framework proved
to be just as useful for nonvisual objects and is nowadays frequently used as the
basis for other frameworks—including Tapestry.
The component approach has been widely used in many environments to
support traditional GUI development, including Java’s AWT and Swing. Using a
component approach for web applications is less widespread but is not original
to Tapestry. As far back as 1994, NeXT’s WebObjects framework integrated a
component object model with graphical editor tools and an object relational
database layer (to handle long-term data persistence, much like entity Enterprises JavaBeans in a Java 2 Enterprise Edition application).
In WebObjects, components were responsible for both generating portions of
an HTML page and responding to user requests, and the framework was responsible for hiding the details of HTTP and server-side state management. This
approach to componentizing web application behavior matches the high-level
design goals of Tapestry, though the overall design and implementation of the
two frameworks are completely different.
1.3.3 What is Tapestry?
The Tapestry framework is a layer between a Java servlet container and a Tapestry application. Tapestry is not a standalone server; it is an extension to servlets
and works within existing servlet containers (such as Tomcat) or application servers (like JBoss, WebSphere, or WebLogic), which include a servlet container. Figure 1.4 shows how Tapestry fits into the overall picture; the application consists
of pages, which are themselves constructed from reusable components. The
application operates within the application server. To the application server, the
application is just another servlet.
Web applications are typically implemented in terms of three layers, each
addressing a different concern within the overall application:
19
Figure 1.4 Tapestry applications run within a servlet container or application server; inside
the server, the application consists of individual pages built from components. The Tapestry
application servlet links the requests from the servlet container to the pages and
components of the application.
■
The presentation layer is responsible for receiving incoming HTTP requests and
forming HTML responses.
■
The application layer is responsible for all business logic; this is often implemented using Enterprise JavaBeans (EJBs).
■
The database layer is responsible for storing data persistently.
Tapestry is only the presentation layer of an application. By this, we mean that
Tapestry deals only with presenting information to an end user, as HTML,8 and
8
Actually, very little of Tapestry is HTML specific. Tapestry is equally at home creating Extensible
Markup Language (XML), Extensible Hypertext Markup Language (XHTML), or Wireless Markup
Language (WML). A suite of WML-related components is included with the framework.
20
CHAPTER 1
handling input from the user in the form of clickable links and HTML forms.
Tapestry doesn’t know, or care, about the application layer, the domain logic, or
the source of information. A Tapestry application can be built on top of anything
from a flat-file database, to a relational database accessed via Java Database Connectivity (JDBC), all the way up to a globe-spanning network of EJBs. Tapestry
doesn’t mandate anything about these aspects of your application; it is concerned solely with how the presentation layer is organized.
The Tapestry framework fills a gap that is often missing in web application
development: It provides a consistent structure, a skeleton for developers to work
with. Too often, each developer is left to manage on his or her own, a freedom
that is missing from other engineering disciplines and one that entails a high
degree of risk. Imagine building a skyscraper by dividing the job into constructing individual floors and instructing each worker to design and build a single
floor, with the intention of using a crane to stack them all up at the end. Each
worker is provided with a rough sketch of the finished building and told what
kind of work will be done on that floor: “We’ll be putting the accounts payable
department here,” or “This is where the company cafeteria will go.” You can
imagine the kind of disaster that would result; free of any constraints, each
worker would use different materials and a different layout. Elevator shafts
wouldn’t line up, some workers would forget to leave room for electrical wiring
or plumbing…and some might even forget to put in windows!
And yet many web applications are developed under a similar kind of chaos.
Specifications are incomplete; different developers approach similar problems in
different ways. Some developers are unaware of important details of the project
or of the many minor details of web application development. Skills sets vary, so
some developers excel at writing client-side JavaScript, others are good at interacting with back-end systems—but project management may treat the developers as completely interchangeable. Too often, the success or failure of entire
projects rests on the shoulders of a “SWAT team” that must glue all the bits and
pieces of code together at the last minute.
Continuing with the skyscraper metaphor, Tapestry changes the picture significantly. Instead of starting with individual two-by-fours, the workers start with
the empty shell of each floor, with most of the plumbing and wiring already in
place. It just becomes a matter of installing those fixtures and features unique to
that floor, all with the knowledge that everything will stack up properly at the
end of the project. Translate floor to page, fixture to component, and install to configure, and you’re well on your way to understanding what Tapestry offers you as
a developer.
21
Tapestry is not a panacea for all development problems; however, it does provide a consistent way to describe the implementation of a web application project
and the interaction between elements of the project provided by different developers. Web applications don’t have elevator shafts, but they do have, for example,
consistent navigation bars, or embedded search forms, or login buttons. These
elements can be implemented as Tapestry components, tested separately, and
reused on every page, thus guaranteeing both a consistent visual look and a consistent interactive behavior.
1.3.4 Comparing Tapestry to Swing
Tapestry represents a different way of assembling web applications, one based on
combining and configuring components. Tapestry development is much closer
in style to creating an application using the Swing GUI framework than traditional web application development with servlets and JSPs.
Developers who are familiar with creating desktop applications (using Swing
or AWT or even toolkits for other languages and environments) are often left in
the lurch when transitioning over to web applications. The familiar patterns of
development don’t translate well. For desktop applications, you expect to combine and configure existing components to create your application. You expect
individual components to encapsulate both output (what gets drawn to the
screen) and input (what happens when the user types or clicks the mouse). The
components, and the framework they build upon, shield the developer from the
minute details of the desktop environment: It is rare that you need to be concerned with event queues or pixel-oriented screen updates when developing a
Swing application because the JButton component already knows how to draw
on the screen and respond to user input. Traditional servlet development
doesn’t contain anything similar; servlets and JSPs are always custom written and
tied directly to the overall application.
Tapestry restores this style of development. Tapestry pages consist of components that are configured and connected to one another. The framework allows
you to create components, pages, and applications without any awareness of servlets. Like Swing developers, Tapestry developers know objects such as HttpSession,
HttpServletRequest, and HttpServlet are present; there just isn’t any pressing
need to access those objects—that’s Tapestry’s responsibility, encapsulated within
the framework and the Tapestry components. Tapestry performs all the necessary dispatching of incoming requests and uses an event-notification system to
get application-specific code executed.
22
CHAPTER 1
1.4 Understanding Tapestry’s goals
At the core of Tapestry’s design is a vision: It should be easier for you to do the
right thing than the wrong thing—Tapestry should allow you to avoid the pitfalls
and antipatterns prevalent in traditional web application development. You
should end up with a robust, scalable, maintainable application not because the
framework merely allows it, but because the framework makes that the clearest,
simplest option to follow. This vision is expressed in terms of four goals that
influence the design and implementation of the framework: simplicity, consistency, efficiency, and feedback.
1.4.1 Simplicity
Tapestry applications contain a surprisingly small amount of Java code. The
stateless, operation-centric programming model used by normal servlet applications requires far too much Java code—code to extract and interpret query
parameters from the request, code to manage data stored as session attributes,
and so on—code for all the uninteresting plumbing. At the center of all this
plumbing is the small amount of application-specific logic: what to do with the
data once it has been extracted, translated, converted, and validated. In a Tapestry application, the Java code you write is just that application-specific logic (in
the form of a listener method, as you’ll see in chapter 2). Combining a generic
component with the application-specific listener method is a simple and elegant
way to structure an application. All that plumbing is no longer your responsibility; it’s all buried inside the Tapestry framework.
You tell the framework what needs doing, and it takes care of the details. For
example, rather than write code to build and format a URL with query parameters, you just use an existing component and configure it to execute a method
you supply when it is triggered. Likewise, Tapestry’s HTML form support takes
care of reading object properties when a page is rendered, but also reads and
interprets query parameters and updates object properties when a form is submitted (this is covered in chapters 3, 4, and 5).
1.4.2 Consistency
Consistency on a large team project can be a godsend. If you’ve ever been
brought into an existing project, or have even filled in temporarily for another
team member on a project, you’ve likely experienced “impedance frustration.”
That is, each developer names things a little differently, solves problems a little differently, puts code in different places, and so forth. Before you can be productive
Understanding Tapestry’s goals
23
in another developer’s code, you have to figure out that developer’s style, and
that causes frustration.
With Tapestry, you get a consistent environment and approach across pages,
even when different pages are the responsibility of different developers. There’s
little or no guesswork when adding new links or forms to pages; they all work
pretty much the same way because they are based on the same reusable components. Tapestry makes it easy to create new components for common functionality
in an application. The application behaves consistently and is coded consistently
because of reuse of both components and code.
1.4.3 Efficiency
It is important that Tapestry applications be scalable; there’s no point in creating a
web application if it can’t handle a reasonable number of concurrent users on reasonable hardware. Internally, Tapestry uses object pools and caches to minimize
the amount of processing that must occur during a request. For example, Tapestry
will read each XML specification file and each HTML template exactly once and
store the file’s contents for later use within the application. In addition, the framework is structured so that expensive application operations (for example, performing a Java Naming and Directory Interface [JNDI] lookup to resolve an EJB’s home
interface) can be performed once and cached for fast access when needed again
elsewhere within the application. You’ll see examples of this in chapter 10.
Tapestry applications have been compared to equivalent servlet or Struts
applications and found to have similar performance curves. Such results match
the traditional wisdom that the presentation layer is rarely the application performance bottleneck; the time it takes to process a request and render a response
is usually gated by the speed with which data can be obtained from the application’s back-end database.
1.4.4 Feedback
In most web application frameworks, when something goes wrong in your code
or in the framework code, you will see a stack trace in the web browser. Suddenly,
you are forced to play detective, working backward from the cryptic clues provided in the stack trace to the problem. Does your code contain an error? Is
there a typo in a deployment descriptor? An error communicating with your
application server? A deployment problem with an EJB? You can waste large
amounts of valuable developer time tracking down often-trivial problems.
The main code path into a Tapestry application has multiple layers of exception catching and reporting that ensure that a reasonable exception report is
24
CHAPTER 1
Figure 1.5 The Tapestry exception report page starts by identifying the nested exceptions, providing
a stack trace for the most deeply nested exception. The exception also identifies the file and line
associated with the error. In this case, line 27 of file Home.html had a typo: listenersstart instead of
listeners.start.
displayed either in the client web browser or in the server’s console. Figure 1.5
shows the initial portion of such an exception report (this exception was provoked by introducing a typo into the HTML template for the Hangman application, described in detail in chapter 2).
25
Tapestry’s approach is to provide you with as much information as possible
to help you rapidly fix the problem. This is represented by five factors visible
in figure 1.5:
■
Tapestry has worked through the stack of exceptions, starting with org.
apache.tapestry.BindingException and digging down to ognl.NoSuchPropertyException.
■
For each exception, it has displayed the exception message and all the
properties of the exception (binding and location for the BindingException,
and target for NoSuchPropertyException).
■
The report identifies the exact file and line that is in error—the location
property of the BindingException indicates that the error is on line 27 of
file Home.html.
■
The stack trace for the deepest exception, NoSuchPropertyException, is the
most relevant, and that’s the only one displayed.
■
Tapestry has attempted to describe in the exception message exactly what
went wrong.
Stack traces and exceptions are not always enough. Sometimes to understand a
problem you need to know more about the request and general environment. As
shown in figure 1.6, the exception report continues with exhaustive output
about the key Servlet API objects (HttpServletRequest, HttpSession, HttpServlet, and HttpServletContext), followed by a listing of all Java Virtual Machine
(JVM) system properties. Collecting this kind of information would normally
require restarting the application and using the Java debugger. Tapestry saves
you time by providing all this useful information immediately, at the time of the
initial error, which means less time tracking down bugs and more time for
everything else.
With these four central goals in mind, it’s now time to start describing, at a
high level, how Tapestry operates.
1.5 How does Tapestry work?
Tapestry is an extension that builds on the standard Java Servlet API. Tapestry
still uses a servlet and still interacts with all the Servlet API objects: HttpServletRequest, HttpSession, and so forth. Tapestry applications are still servlet applications and can be deployed into any standard servlet container, such as Apache
Tomcat, Jetty, Resin, WebLogic, or WebSphere. The framework is code compatible
26
CHAPTER 1
Figure 1.6 The exception report includes a vast amount of information about the Request, Session,
Servlet, ServletContext, and all JVM system properties—the kind of information you would normally
need to collect using the Java debugger.
with Java Development Kit (JDK) 1.2 and Servlet API 2.2,9 which allows Tapestry
to be used in virtually any recent Java application server.
9
Tapestry includes a servlet filter that performs a redirect of the initial application request to the Tapestry application servlet; servlet filters are a feature of Servlet API 2.3. Chapter 10 discusses the need
for this filter and how to make use of it.
27
Tapestry applications consist of any number of pages, where the pages are
each constructed from individual, reusable, configurable components. Table 1.1
defines the terms used to describe Tapestry applications.
Table 1.1
Basic terms used when describing Tapestry applications
Term
Description
Page
Applications consist of a collection of uniquely named pages; each page has a template and contains other components.
Template
An HTML template for a page (or a component). In Tapestry, a template contains ordinary HTML markup, with certain tags marked with a special attribute to indicate they
are placeholders for components.
Component
A reusable object that may be used as part of a Tapestry page. Components generate
HTML when a page is rendered and may also participate when a link or form in the
rendered page is triggered. Components may also be used to construct new components.
Parameter
Components have parameters, which link properties of the component to properties of
the page (or of domain objects accessible from the page). Components mostly read
their parameters, but some components (often related to HTML forms) can update
their parameters and thus update the page properties bound to the parameter.
Your Tapestry application includes templates and specifications for these elements. Those elements work together to form your GUI.
1.5.1 What’s in a Tapestry application?
Like other servlet applications, Tapestry applications are most commonly distributed as Web archives (WARs). A WAR is a variation of an ordinary Java archive
(JAR) file. A JAR file is used to store Java class files, but a WAR file stores a mix of
files: Primarily it stores HTML files, stylesheets, and images that will be directly
accessible to a client web browser using a URL. Like a JAR file, a WAR file contains
many files organized into many folders within the WAR.
A deployed web application is also called a web application context. The
default name of the context is based on the name of the WAR . Figure 1.7 diagrams a deployed web application, the Virtual Library application described in
chapters 9 and 10. The Virtual Library is deployed as vlib.war, which will
define a web application context URL as /vlib. Static files (those images and
stylesheets) are accessible at URLs that extend from the context URL, such as
http://server/vlib/images/search.png (which accesses a file stored as images/
search.png inside vlib.war).
28
CHAPTER 1
Figure 1.7 The WAR file, vlib.war, defines a context, /vlib. Resources inside the WAR file can be
referenced by extending the URL for the context itself.
WARs also contain a WEB-INF folder; this includes private resources that are not
made visible to clients. A servlet application will store class files and libraries
inside the WEB-INF folder. Tapestry applications go further, storing additional
artifacts and resources in the WEB-INF folder:
■
The compiled Java classes are stored in the WEB-INF/classes folder.
■
The WEB-INF/lib folder stores JAR files needed by the application’s classes.
In a Tapestry application, the WEB-INF/lib folder will contain the Tapestry
framework JAR (e.g., tapestry-3.0.jar) as well as a number of supporting
libraries needed by the framework.
■
Page specifications are stored in the WEB-INF folder, with a .page extension.
■
Page templates are stored in the root context folder (that is, directly within
vlib.war) with an .html extension.
■
Component specifications are stored in the WEB-INF folder with a .jwc
extension.
■
Component templates are stored in the WEB-INF folder with an .html
extension.
■
The WEB-INF folder also contains the web deployment descriptor, web.xml.
This file defines, to the servlet container, all of the servlets and URL mappings contained in the WAR. A Tapestry application uses only a single servlet but still requires a deployment descriptor.
29
Appendix B discusses the example web applications for this book; it includes
details on how to build and deploy the examples on your own workstation. It also
goes into some detail about the directory layout for the examples, and how the
Ant build tool is used to compile, package, and deploy the examples.
1.5.2 Tapestry’s Model-View-Controller (MVC) pattern
Perhaps the most successful design pattern in GUI development is the ModelView-Controller pattern (MVC). Most GUI frameworks, including Swing, use
some variation of MVC. The MVC pattern divides up each user interaction within
the application into three categories of objects:
■
■
■
Model—The Model object is used to store information that will ultimately
be presented to the user (and possibly edited). These objects are often
domain objects—objects that represent the specific application domain,
such as a Customer, Order, or Product (in an e-commerce application).
The Model should be completely independent of the GUI. In Java applications, the Model is often a JavaBean.
View—The View object is responsible for presenting data obtained from
the model in a format appropriate to the application. In a web application,
a simple View may do no more than write a string into the rendered page;
a very complex View may create a graphic chart based on data in the
Model, and an editable View may write a text field or other form element
into the response.
Controller—The Controller object has two functions: First, it bridges
between the Model and the View, reading data from the Model and providing it to the View. Second, it is responsible for interpreting user input and
updating the Model in response; in a web application, the Controller will
service incoming requests (including form submissions).
As shown in figure 1.8, the Model and View have no direct connection; instead,
the Controller mediates between the two and has the added responsibility of processing user input.
The MVC pattern is useful because it emphasizes separation of concerns, a
powerful concept that supports more flexible, more robust development. The
Model and the View are kept entirely separate; they may be developed at different times, by different developers, and tested independently. The Model
has no knowledge that it is part of a GUI or any type of application. The Controller is responsible for monitoring the Model and informing the View of relevant changes when the Model changes. Typically, the Model includes an
30
CHAPTER 1
Figure 1.8
Using the Model-ViewController pattern separates
the View and the Model,
which have no direct
knowledge of each other.
The Controller mediates
between the two and
handles user input.
event mechanism so that any changes will be propagated, through the Controller, to the View. A single Model may be represented by multiple Views,
either as different options or simultaneously in multiple windows. A common
example is to display tabular data as both a table and a chart—that’s two different Views of the same Model. The MVC design pattern has been battletested in countless applications.
Using servlets as Controllers
Within the realm of typical servlet applications, the MVC pattern typically
takes on the form shown in figure 1.9. In a servlet application, servlets play
the role of Controller. A servlet will receive a request from the client and will
perform an operation on the Model, the domain-specific objects for the application. For example, the request may be a form submission, and the Controller servlet will update properties of the Model domain objects and store them
into a database.
The Controller servlet will then load up the HttpServletRequest with
attributes needed by the View to render a response, in effect “pushing” information from the Model into the View. The View itself will be a JSP (or some other
form of template) and can draw information from the domain objects (the
Model) as it renders the response to be returned to the client.
Tapestry makes use of a common variant of MVC, where the View “pulls”
information out of the Model (rather than having the Controller push information into the View). The View still has no explicit knowledge of the application; it
31
Figure 1.9 The servlet, as Controller, receives the request. It locates and updates the domain
objects, possibly reading and updating database data. The Controller servlet selects a View (a JSP)
to render the response. The View draws data from the domain objects as it renders, and the final
response page is sent to the client.
just knows where it can read (and occasionally write) data. The Controller is still
responsible for handling user input, as well as for establishing the relationships
between Model and View.
The component as Controller
As shown in figure 1.10, a Tapestry component fills the role of Controller,
mediating between pure-domain objects in the Model (the LineItem and Product objects, in this example) and components contained within its HTML template (TextField and Insert). Most often, this pattern applies to pages (pages
are still Tapestry components), but in many cases, a component will have its
own template, containing further components, and support its own interactions with the user.
The page establishes relationships between the Model and the View in terms
of property expressions. Property expressions are implemented using another open
source framework, the Object Graph Navigation Language (OGNL), a very powerful Java expression language.
The page exposes the Model objects to the View components by providing
JavaBeans properties (such as the lineItem property in figure 1.10)—but
that’s just a starting point. Properties of the Model objects are bound to the
components—when the page renders, the TextField component will read the
quantity property of the LineItem object, and the Insert component will read
the name property of the Product object. This approach of navigating the
32
CHAPTER 1
Figure 1.10 Model-View-Controller in Tapestry: The page is the Controller, LineItem
and Product are domain objects forming the Model, and the TextField and Insert
components are the View. The TextField component is bound to the quantity property
of the LineItem object (which it both reads and updates), and the Insert component is
bound to the name property of the Product object.
object graph to reach relevant properties is possible only because the page
and all the components are stateful objects—JavaBeans with properties that
can be read and updated. The page, in its role as Controller, is responsible for
making these connections between the domain objects (the Model) and the
components (the View).
OGNL
33
OGNL is a separate open source project distributed with, and heavily utilized by, Tapestry. OGNL’s primary purpose is to read and update JavaBeans properties of objects. A basic form of OGNL expression is a chain of
property names, separated with periods, as in lineItem.product.name.
This OGNL expression is roughly equivalent to the Java expression getLineItem().getProduct().getName() or getLineItem().getProduct().setName(), depending on whether the expression is being used
to read or update a property. However, that’s just the start of what
OGNL can do. It is an extremely powerful expression language, mimick-
ing (and in some cases exceeding) the capabilities of Java language expressions. It predates the expression language introduced with the Java
Standard Tag Library (JSTL) and is both easier to use and more powerful. OGNL is the creation of Drew Davidson and Luke Blanshard. Full
documentation for OGNL, as well as source and binary downloads, is
available at http://www.ognl.org.
The page also fulfills its role as the Controller by providing the logic that occurs
when links within the page are clicked, or when forms within the page are submitted. This logic is provided in the form of listener methods, short methods
implemented in the page class that are invoked by the Tapestry framework.
1.5.3 Tapestry classes
The Tapestry framework consists of over 400 classes and interfaces—fortunately, when building your own applications with Tapestry, you’ll most often
need to be concerned only with the handful of classes, interfaces, and methods
shown in figure 1.11.
At the root of the hierarchy is the IRender interface,10 implemented by all
objects that can render HTML. Components are the main objects that can render, but not the only ones: for example, static HTML from a page or component
template is wrapped up as an object that implements the IRender interface.
Two key interfaces: IComponent and IPage
Two key interfaces are IComponent and IPage, which define Tapestry components
and pages, respectively. All Tapestry code is coded against interfaces, not implementations; therefore, IComponent is consistently used throughout the framework
as a parameter type or return value, never AbstractComponent, an implementation
10
All Tapestry interfaces start with a leading I, as in IRender or IComponent. The only exceptions are
event interfaces, such PageDetachListener. These are named as such for compatibility with JavaBeans naming conventions for event listener interfaces.
34
CHAPTER 1
Figure 1.11 Here are the key interfaces, classes, and methods provided in the
Tapestry framework.
of IComponent. The AbstractComponent class is the base class for implementing
components. AbstractComponent is an abstract class; it defines but does not
implement the renderComponent() method. Subclasses of AbstractComponent
implement this method, producing all HTML output in Java code.
Using Spindle
35
BaseComponent extends AbstractComponent, adding initialization logic that
knows how to locate and read a template. Most components you create will subclass BaseComponent. We discuss creating components in chapters 6 and 8.
In Tapestry, pages are specialized components with additional responsibilities; this is shown by having IPage extend IComponent, and by having BasePage
subclass BaseComponent. When creating new pages, you will always create subclasses of BasePage.
Three useful interfaces: IRequestCycle, IMarkupWriter, and IEngine
Most Tapestry pages and components can be coded with references to just three
additional interfaces:
■
IRequestCycle—A request cycle stores information about the current
request. It tracks the active page involved in the request and is used to
orchestrate the rendering of a response. The request cycle may also be
used to access the Servlet API objects (HttpServletRequest, HttpSession,
and HttpServletResponse) in the rare event that such access is necessary.
■
IMarkupWriter—A markup writer is used to produce HTML output when a
page is rendering a response. It operates much like a java.io.PrintWriter
but includes additional methods useful for creating markup output (containing XML-style elements and attributes).
■
IEngine—The engine is a central object with several related responsibili-
ties, the core around which a Tapestry application hangs. Primarily, the
engine is responsible for maintaining server-side state, but it also acts as a
gateway to a number of subsystems used internally by Tapestry.
Most of the examples in the first few chapters are limited to just these interfaces;
as we get under the hood in chapters 6 through 8, a few additional interfaces will
be introduced.
1.6 Using Spindle
In this book, we’ll be sticking close to the source with our many example Tapestry
applications, presenting the different artifacts (the HTML templates, the XML
specifications, and the Java code) in listings. If your development environment of
choice is Eclipse (http://www.eclipse.org), you should check out Spindle, the Tapestry plug-in for Eclipse. Spindle blends customized wizards and editors for Tapestry
applications into the Eclipse IDE. Spindle streamlines Tapestry application development in a number of ways:
36
CHAPTER 1
Figure 1.12 Using Spindle, errors in your templates and specifications are highlighted in the
editor and listed in the Tasks view.
■
Spindle scans all types of input files for a wide variety of errors, including
invalid OGNL expressions, unresolvable class names, invalid or unknown
component IDs, and so forth.
■
Errors are identified in the Eclipse Tasks view.
■
Errors are indicated with markers in the editor gutter.
Figure 1.12 shows an example, where some errors have been introduced into
part of the example J2EE application described in chapters 9 and 10.
Spindle takes Tapestry to an even higher level of productivity, because it can
catch errors at build time that ordinarily aren’t caught until runtime. Like Tapestry, Spindle is distributed as a free, open source project. Spindle is available at
http://sf.net/projects/spindle and was developed by Geoff Longman.
Summary
37
1.7 Summary
Ordinary Java web applications are code heavy: Adding a new interaction (such as
a link or form) to a page requires creating new classes—new servlets (or new Struts
Actions) to handle the interaction. Quite a bit of cookie-cutter code must be written just to establish the context of the interaction—which server-side objects
should be affected and how. You, the developer, are responsible for making a series
of small decisions: what to name the new servlet, what URL pattern to map to the
servlet, what to name any query parameters, and what information to store in
those parameters. Additionally, you must make coordinated changes to Java code,
to the JSP, and to the web deployment descriptor. This is all unavoidable, simply
an offshoot of the operation-centric approach mandated by the use of servlets, and
the way in which servlets within an application are weakly bound to one another.
Tapestry applications flip this situation on its ear; a Tapestry application consists of a tiny amount of application-specific code within a web of objects, methods, and properties, and uses XML specifications and OGNL expressions, not
Java code, to tie it all together. Tapestry extends your reach as a developer
because of all the decisions you don’t have to make and all the code you don’t
have to write and test. You get to concentrate on the critical aspects of your own
application without having to address all the generic concerns of structuring a
servlet application.
In the next chapter, we’ll put together a small Tapestry application to demonstrate concretely how the framework’s component-based approach to development really does simplify and accelerate application development.
Getting started
with Tapestry
This chapter covers
■
Creating HTML templates, page specifications,
and page classes
■
Using Tapestry components inside an HTML
template
■
Creating clickable links
■
Encoding extra information into link URLs
■
Configuring Tapestry applications for deployment
38
Introducing the Hangman application
39
In the first chapter, we made a number of claims about what Tapestry is capable of;
now it is time to start backing up those claims with hard code. Launching into a
complete Java 2 Enterprise Edition (J2EE) application right here would be a bit
premature; instead, we’ll start with more of a toy, an application that plays the simple word game Hangman.† In effect, Hangman is a “scale model” of a real Tapestry
application; it demonstrates the basic capabilities of the framework and will give
you an initial sense for what developing Tapestry applications is all about. Along
the way, you’ll see how to:
■
■
■
■
■
■
Separate business logic from presentation logic, within the Model-ViewController (MVC) pattern (described in chapter 1)
Combine HTML templates, page specifications (in XML), and Java classes
to form pages within the application
Create HTML hyperlinks that activate application logic when clicked
Encode custom application data into HTML hyperlinks
Manage server-side state information
Configure a Tapestry application for deployment inside a servlet container
More importantly, you’ll see quite a bit about the work you don’t have to do,
because the framework takes care of it for you.
Appendix B covers how to obtain the source code for all the examples in the
book, as well as how to build the examples on your own computer and deploy
them into the Tomcat servlet container (Tomcat is an open source servlet container available from http://jakarta.apache.org/tomcat). Once Tomcat is running
and you have downloaded the source code, you can launch the Hangman application by opening a web browser to http://localhost:8080/hangman1/app.
2.1 Introducing the Hangman application
Hangman is a simple word game for two players, played on a piece of paper or
on a chalkboard. One player selects a secret target word; the other player attempts
to guess the word. To start, you draw an empty gallows. The guessing player
selects a letter from the alphabet; if the letter appears in the target word, the
other player writes the letter in each position of the target word that the letter
appears in. Each unsuccessful guess is marked by adding a line to a stick figure
†
An even simpler example of a “Hello World” Tapestry application is available at http://www.manning.
com/lewisship/helloworld. The helloworld.war file is pre-compiled and pre-built, containing all the
necessary Tapestry libraries and deployment descriptors. It may be downloaded directly into your
servlet container, Tomcat or otherwise, and accessed as http://localhost:8080/helloworld/app.
40
CHAPTER 2
Figure 2.1 A Tapestry Hangman game in progress. The player has successfully
guessed the letter A, but has also guessed E, P, and V, which are not in the target
word. An important aspect of this application is the look and feel, which should
resemble a game played by hand on a chalkboard.
on the gallows: the head, torso, and then the limbs. The game is over when the
word is guessed or the stick figure is completed.
The Tapestry Hangman application captures all of this functionality and, at
the same time, attempts to capture the classic look of playing the game by hand
on a chalkboard. The user interface makes use of images to represent the letters
and other artifacts of the game, to provide a “hand–scrawled” look and feel. Figure 2.1 shows the middle of a game of Hangman; the player has made several
wrong guesses, so parts of the stick figure are filled in, and one letter (A) has
been guessed correctly so far.
At this point, all we have is a general idea for the application; before we can
get to the coding stage, we must formalize this general idea into something a bit
more concrete—and that begins with identifying the application flow.
41
2.1.1 Determining the application flow
The application flow is a model of how the end user will navigate through the
application. Determining the flow occurs very early in the development cycle; it
is driven by the specific requirements and use cases of the application. Application flow is the most abstract model of the application; it identifies the different
pages in the application and how they are connected, but rarely has to precisely
identify what is on any particular page. Key aspects of the application user interface are discernable from the flow diagrams, such as the need for common navigation menus or specific links between individual pages.
The flow of the Hangman application is quite simple: From the Guess page,
the user makes guesses at the target word, eventually winning or losing the
game. Figure 2.2 is a state diagram for this simple application; when the application is launched, the user is presented with a Start page (figure 2.4); from there,
he or she can start a new game, making guesses that eventually reach either a
win or a loss; from there, the player can restart the game with a new target word.
From this simple description, you can see that we’ll have four distinct pages in
the application:
Figure 2.2
The player starts the
game and makes
guesses, eventually
reaching the win or lose
page, from which the
player can start a new
game (with a new word).
42
CHAPTER 2
■
Start—A welcome page to greet players before starting a new game
■
Make Guess—The main page, from which players may guess letters of the
target word
■
Win—The page reached after the target word is successfully guessed
■
Lose—The page reached after players have exhausted their guesses
Once the application flow has been determined, the next step is to prototype
what the individual HTML pages will look like.
2.1.2 Creating page mockups
Page mockups are static HTML pages that represent what the active pages from
the running application will look like. These are ordinary HTML pages with
placeholder values representing the content that will eventually be generated
dynamically by the application. The point of creating the mockups is to give
the HTML developers a chance to work out the look and feel of the application,
right down to fonts, colors, and graphics, without concern for how the application will be implemented.
Figure 2.3 shows the mockup for the Guess page in an HTML editor. The
HTML source is shown in the upper pane, and the WYSIWYG preview appears in
the lower pane. This mockup will eventually be converted for use as the Guess
page’s HTML template.
Page mockups should display all the features of the running application,
especially such features as error messages that are included only conditionally.
For example, a mockup may include a snippet for an error message:
<img src="images/error.png" width="32" height="32" align="top"/>
<span class="error-message">
Placeholder for error message.
</span>
This snippet is important for two reasons: It clearly identifies how a real error
message should be displayed, and it identifies exactly where within the page the
error message should be displayed. In the Guess page mockup, the Guess and
Choose sections demonstrate what the page looks like in the middle of a game,
with some letters of the target word filled in and several letters from the alphabet
already guessed. Having clear examples of these dynamic aspects of the page will
be invaluable to the Java developer when he or she is converting the mockup
into a usable HTML template.
It is not an absolute requirement that you create a mockup for every page in
the application; often, mockups for only a handful of key pages will suffice, and
Figure 2.3
43
The page mockup for the Guess page in an HTML editor.
developers can use these core mockups as templates for the remaining application pages.
As you’ll see shortly, converting these HTML mockup pages into usable Tapestry page templates requires a minimal number of unobtrusive changes. A
mockup is converted into a page template by adding instrumentation: Additional
tags and tag attributes are used to identify and configure Tapestry components
within the template. This instrumentation is designed to be nearly invisible. Tapestry’s approach stands in stark contrast to the use of JSPs, where the conversion
44
CHAPTER 2
from HTML mockup to JavaServer Page (JSP) is a one-way process. Once the
HTML mockup page has been converted to a JSP, it will not preview correctly in a
standard HTML editor, making all subsequent changes to the JSP that much
more difficult. Within Tapestry, a page template can still be edited by an HTML
developer using standard HTML editing tools; in effect, the mockups evolve into
the HTML page templates yet can still be treated as mockups.1
This is an important aspect of Tapestry because late changes to application
flow and look and feel are simply a reality when creating web applications—
there’s always a last-minute change: a new page to add, a background color to
change, or a column width to tweak. Even in an impossibly idealized project, one
where no late changes ever occurred, a subsequent release of the application
would inevitably update the application flow and at least some aspect of the look
and feel. Tapestry accommodates these kind of late cycle changes quite well
because of how unobtrusive the instrumentation (the additional tags and tag
attributes used to identify components within a template) is. Much more work
can be done by an HTML developer using standard HTML editing tools, without
the involvement of Java developers.
Meanwhile, even as the HTML developers are working on the mockups, the
Java developers should be getting a head start on the design of the actual application, and that begins with identifying the domain objects.
2.1.3 Defining the domain objects
The architects and developers on the Java side of the team are ultimately
responsible for the running application; in most applications, this becomes a
question of linking a user interface to your domain objects. Domain objects are the
objects of the middle tier, the application tier, in the overall application—they
are the entity objects for data stored in a database, or objects that implement
your business’s specific processes. Common problems to solve involve what information is stored by these objects, how the different objects are related, and how
they are read from, or stored into, a database.
Even in a simple application such as Hangman, which does not make use of a
database, there are still domain objects, and still advantages (in accordance with
the MVC design pattern) to keeping these objects well separated from any code
directly related to the user interface.
1
Tapestry isn’t magic, and there are some limitations on maintaining full WYSIWYG previews of
HTML templates once more sophisticated custom components are created and used within a page
template; this subject is covered in chapter 6.
45
The two domain objects used in the Hangman application are WordSource
and Game. The first, WordSource, is simply a wrapper around a list of words read
from a text file and is used to dole out random words for the player to guess.
Game is a bit more interesting; it encompasses all the logic about the game. Specifically, the Game object knows:
■
The target word the player is attempting to guess
■
Which of the 26 letters of the alphabet the player has already guessed
■
Which letters of the target word have been filled in by successful guesses
■
How many incorrect guesses remain
■
If the player has won the game (by guessing all the letters in the target word)
As promised, the implementation of the Game class (in listing 2.1) knows nothing
about Tapestry or any other user interface.
Listing 2.1 Game.java: domain object for the Hangman application
package hangman1;
public class Game
{
private String _targetWord;
private int _incorrectGuessesLeft;
private char[] _letters;
private boolean[] _guessed = new boolean[26];
private boolean _win;
public boolean isWin()
{
return _win;
}
public char[] getLetters()
{
return _letters;
}
Returns true
once word has
been guessed
Returns array
of letters in
the word
public int getIncorrectGuessesLeft()
{
return _incorrectGuessesLeft;
}
public boolean[] getGuessedLetters()
{
return _guessed;
}
Returns 26 flags:
letters guessed
by player
46
CHAPTER 2
public void start(String word)
{
_targetWord = word;
_incorrectGuessesLeft = 5;
_win = false;
int count = word.length();
_letters = new char[count];
Starts
a new
game
for (int i = 0; i < count; i++)
_letters[i] = '_';
for (int i = 0; i < 26; i++)
_guessed[i] = false;
}
public boolean makeGuess(char letter)
{
char ch = Character.toLowerCase(letter);
if (ch < 'a' || ch > 'z')
throw new IllegalArgumentException(
"Must provide an alphabetic character.");
int index = ch - 'a';
if (_guessed[index])
return true;
_guessed[index] = true;
boolean good = false;
boolean complete = true;
for (int i = 0; i < _letters.length; i++)
{
if (_letters[i] != '_')
continue;
Processes
a player’s
guess
if (_targetWord.charAt(i) == ch)
{
good = true;
_letters[i] = ch;
continue;
}
complete = false;
}
if (good)
47
{
_win = complete;
return !complete;
}
if (_incorrectGuessesLeft == 0)
{
_letters = _targetWord.toCharArray();
Processes
a player’s
guess
return false;
}
_incorrectGuessesLeft--;
return true;
}
}
The makeGuess() method is invoked to process a player’s guess. It updates the
target word and other properties and returns true if more guesses are allowed. It
returns false if the player has either won or lost the game.
The Game class must provide some support for the user interface, but it does
so in a generic fashion without being tied to the interface; it’s the Model in the
MVC pattern described in chapter 1. This support takes the form of JavaBeans
properties that are exposed to the user interface, such as the number of incorrect
guesses remaining or the list of letters already guessed. These properties are
bound to Tapestry component parameters, allowing those components to display the number of guesses remaining, the partially guessed word, or the list of
remaining unguessed letters. In addition, Game provides methods that can be
invoked by the user interface code to start a new game or to process a guess
made by the player.
A second class, WordSource, is also used. WordSource is responsible for providing a random word for the player to guess. The source of the words is a small file,
WordList.txt, packaged with the WordSource class. The WordSource class is provided in listing 2.2.
Listing 2.2 WordSource.java: domain object for the Hangman application
package hangman1;
import java.io.IOException;
import java.io.InputStream;
48
CHAPTER 2
import
import
import
import
import
import
java.io.InputStreamReader;
java.io.LineNumberReader;
java.io.Reader;
java.util.ArrayList;
java.util.Collections;
java.util.List;
public class WordSource
{
private int _nextWord;
private List _words = new ArrayList();
public WordSource()
{
readWords();
}
private void readWords()
{
try
{
InputStream in =
getClass().getResourceAsStream("WordList.txt");
Reader r = new InputStreamReader(in);
LineNumberReader lineReader = new LineNumberReader(r);
while (true)
{
String line = lineReader.readLine();
if (line == null)
break;
if (line.startsWith("#"))
continue;
String word = line.trim().toLowerCase();
if (word.length() == 0)
continue;
_words.add(word);
}
lineReader.close();
}
catch (IOException ex)
{
throw new RuntimeException(
"Unable to read list of words from file WordList.txt.",
49
ex);
}
// Randomize the word order
Collections.shuffle(_words);
}
public String nextWord()
{
if (_nextWord >= _words.size())
{
_nextWord = 0;
Collections.shuffle(_words);
}
return (String) _words.get(_nextWord++);
}
}
When WordSource is instantiated, it reads the list of words. Later, the nextWord()
method is invoked to get a new word for the player to guess. The method is
designed to not repeat a target word until every word in the list has been guessed.
As with the Game class, this class has no direct connection to Tapestry—these
objects fit firmly into the Model category within the MVC pattern. This kind of
decoupling from the user interface is very important, because it means the Game
and WordSource classes can be tested without having to run the Tapestry application, which in turn means the code can be fully tested inside an automated test
suite. Making code testable is always a worthy goal, because no matter how simple the code is, when you write tests, you find bugs.
Once all the details of the domain objects are worked out, the next step is to
begin work on the pages that will interact with those domain objects.
2.1.4 Defining the pages
Like any other Tapestry application, the Hangman game consists of a set number of pages, which are themselves composed of components. In a Tapestry
application, each page is constructed by combining three related artifacts: an
HTML template, a page specification, and a Java class.2
2
Refer back to section 1.5.1 to see how to properly package these artifacts for use within a servlet container. Appendix B provides examples of how to set up your development workspace and how to use
Ant to build and deploy the WAR.
50
CHAPTER 2
Each Tapestry page has a specific, unique name. The page name is used to
locate the page specification and HTML template. Part of the page specification
is the name of the Java class to instantiate; this is called the page class, and it will
include properties and methods specific to your application.
The Hangman application contains only four pages: Home, Guess, Lose, and
Win, corresponding to the four pages identified in the application flow state diagram (figure 2.2). The Home page here is the same as the Start page in figure 2.2.
By default, when a Tapestry application is first launched, the framework renders
the page named Home. Although there are several options for changing this
behavior, the simplest approach is to follow Tapestry’s naming convention—by
naming the first page a user will see Home.
Creating a functioning Tapestry page starts with the HTML mockup for the
page. This mockup must be instrumented to act as an HTML template instead of
a mockup. Instrumenting a mockup inserts additional attributes and tags in
the mockup that tell Tapestry which parts of the template are dynamic components. Most of a template, however, is exactly the same as the mockup—simple,
static HTML.
NOTE
In real projects, the mockups are not always available when needed by
the Java developers creating the pages. In this situation, the Java developers will create simple, minimal HTML templates—just enough to wire
up the functionality of the application. When the mockup is ready, some
careful cut and paste from the mockup into the minimal HTML template will convert it to use the desired application look and feel.
Once the HTML template is instrumented, a page specification (a short XML document) can be created. The page specification has a number of responsibilities
(many of which will be discussed in later chapters). Its most basic responsibility is
to identify which Java class is to be instantiated as the page. In chapter 1, we
described Tapestry as being a component object framework; this means that
each component fits into an object hierarchy, either as a container of other components or as a containee of a specific component—or, in many cases, as both
container and containee. Pages are still components, sitting at the root of the
component object hierarchy.
As you’ll see, the page class is specific to the application and contains a mixture of properties and methods that support both the rendering of the page and
any user interaction in the page. Ultimately, the behavior of the page is defined
by the page’s properties and methods, combined with the components contained
51
within the page—including the templates, properties, and methods of those
components. This may seem a bit dizzying in theory, but in practice it all comes
together simply and seamlessly. For our first example, let’s start with the Home
page—the simplest page in the Hangman application.
2.2 Developing the Home page
The Hangman application’s Home page has only one small bit of user interaction: a link that starts a new game. This interaction is triggered by clicking the
Start image, shown in figure 2.4. Like any page, the Home page is a combination
of an XML page specification, an HTML template, and a Java class. Our first
steps into Tapestry will be to examine how these three artifacts are combined to
form a simple page.
Figure 2.4 The Home page of the Tapestry Hangman application. The player may
click the word Start to begin a game.
52
CHAPTER 2
The Home page is displayed when the application is first launched. The Web
archive (WAR) for the application must be deployed into the servlet container,
and the servlet container must itself be running. This WAR will contain the Tapestry framework JARs, the page templates and specifications, the static image
files (and other assets), and the compiled Java classes (this is discussed in
chapter 1, section 1.5.1). When the user launches the application (by opening a
web browser to http://localhost:8080/hangman1/app), the framework responds
by rendering the Home page.
The first step in rendering a page is to create an instance of the page. The
framework reads the Home page’s specification and HTML template and uses
this information to create the page instance. A Tapestry page is not a single
object; the page object is the root of a tree of objects, including Tapestry components from the page’s template, the contents of the HTML template, and a number of objects used to connect the individual pieces together. There’s no special
assembly stage for Tapestry applications, nor are there any special build steps or
compilation—all that is necessary is to package the specifications, templates, and
Java classes inside the WAR.
NOTE
You might be concerned about performance, given all this talk of parsing specifications and templates and instantiating trees of objects—but
don’t be. This parsing occurs very quickly, and, unlike with JSPs, there’s
no time spent compiling generated Java source code (JSP compilation
causes a noticeable delay the first time a JSP is used within a traditional
servlet application). In line with Tapestry’s efficiency goal, all the specifications and templates are read and parsed just once, and then cached
for fast access when needed again in future requests. Page instances are
also stored and reused in later requests.
Let’s dive a little deeper and see exactly how the Home page’s specification is
used by the framework.
2.2.1 Understanding the Home page specification
The framework’s first step toward instantiating the Home page is to locate and
read the page’s specification. Page specifications are validated XML files (with a
.page extension) that are stored in the WEB-INF folder of the web application.
The page specification’s first responsibility is to identify the page class it needs to
instantiate—it has other many other, optional responsibilities that we’ll cover
53
later in this chapter and in subsequent chapters. Listing 2.3 contains the complete specification for the Home page of the Hangman application.
Listing 2.3 Home.page: specification for the Home page
<?xml version="1.0"?>
<!DOCTYPE page-specification PUBLIC
"-//Apache Software Foundation//Tapestry Specification 3.0//EN"
"http://jakarta.apache.org/tapestry/dtd/Tapestry_3_0.dtd">
<page-specification class="hangman1.Home"/>
This is about as simple as a page specification can get; its only purpose is to
identify the page class, hangman1.Home. This is a Java class written for the Hangman application, which will be the runtime representation of the page (see section 2.2.3 for more details). By convention, the class name for a page is the same
as the page’s name (though often stored inside a Java package), but of course,
you are free to ignore this convention and name pages and classes differently. It’s
important, however, that the <!DOCTYPE> declaration be exactly as shown in listing 2.3.
WARNING
Use the correct <!DOCTYPE>. Tapestry uses a validating XML parser to
read specifications. Tapestry is purposely finicky about the public ID
(the first string after PUBLIC), since it uses the known public ID to access
a copy of the document type definition (DTD) inside the framework’s
JAR rather than access it over the Internet using the system ID. The
public ID must exactly match the value in listing 2.3, or an ApplicationRuntimeException is thrown. For example, changing Foundation to
Floundation will result in an exception report with this error message: Document context:/WEB-INF/Home.page has an unexpected public id of ‘-//Apache
Software Floundation//Tapestry Specification 3.0//EN’. Watch out for typos;
this is one area where a little cut and paste will save you some grief.
In addition, there is nothing that keeps a single page class from being used for
multiple pages. Each page will have a distinct instance of the page class, just as
each component in a page is a distinct instance of a component class.
54
CHAPTER 2
2.2.2 Rendering the Home page
After parsing the page specification, Tapestry locates the HTML template for the
Home page. The HTML template, which is named Home.html, is located in the
root of the web application archive. This template is shown in listing 2.4.
Listing 2.4 Home.html: HTML template for the Home page
<html>
<head>
<title>Tapestry Hangman</title>
<link rel="stylesheet" type="text/css" href="css/hangman.css"/>
</head>
<body>
<table>
<tr>
<td><img alt="[Tapestry Hangman]"
src="images/tapestry-hangman.png" width="197" height="50"
border="0"/>
</td>
<td width="70" align="right"><img height="36" alt="5"
src="images/Chalkboard_3x8.png" width="36" border="0"/>
</td>
<td><img alt="Guesses Left" src="images/guesses-left.png"
width="164" height="11" border="0"/>
</td>
</tr>
<tr>
<td>
</td>
<td>
</td>
<td><img alt="" src="images/scaffold.png" border="0"/>
</td>
</tr>
</table>
<br/>
<a href="#" jwcid="@DirectLink"
Dynamic
listener="ognl:listeners.start">
portion of
<img src="images/start.png" width="250" height="23"
template
border="0" alt="Start"/></a>
</body>
</html>
The majority of the HTML template is standard, static HTML; only a single Tapestry extension beyond ordinary HTML is used, showing up in the portion of the
template that provides the link to start the game.
55
The <a> tag declares a Tapestry component within the template, giving us our
first whiff of a dynamic web application rather than a static web page. The attribute
jwcid is the indicator that Tapestry uses to identify components within the template. The name jwcid is simply Java Web Component ID. The component is type
DirectLink, one of over 40 components provided with the Tapestry framework.
The example here is an implicit component, where the type of component and
its configuration are declared directly in the HTML template. The @ symbol
indicates to Tapestry that the component is implicitly declared. Later in this
chapter, we’ll show examples of declared components, which have their type and
configuration stored inside the page specification.
The DirectLink component is used to create a particular type of callback into
the application. This component is one of the two primary ways that interaction
occurs in Tapestry; the other is user-submitted forms (which are covered starting in
chapter 3). The DirectLink component renders an HTML <a> element, supplying a
URL that, when clicked by the end user, causes a specific listener method of the
page to be executed (we’ll discuss what a listener method is shortly, in section 2.2.3).
The position of the DirectLink component within the template is delineated
by the <a> and </a> tags. Everything else in this HTML template is static
HTML—text that is sent through to the client web browser unchanged. Just the
portion rendered by the DirectLink component is dynamic. Figure 2.5 shows
how the dynamic and static portions of the template are integrated together to
form the complete response.
The Home page’s HTML template is divided into five individual “chunks.”
Each chunk is either a block of static HTML, the start tag for a component (recognized by Tapestry because of the presence of a jwcid attribute), or the matching end tag for a component. Chunk b is the portion of the HTML template
that precedes the DirectLink component. Chunk c is the component itself.
Chunk d is the portion of the page enclosed by the DirectLink. Chunk e is the
close tag for the DirectLink component. Chunk f is the remainder of the template after the DirectLink.
Chunks that are enclosed directly within a component’s start and end tags are
part of that component’s body. This is a very important part of Tapestry: Components control if and when their bodies are rendered. We’ll frequently refer to
the body of the component: This is the static HTML and other components that
are enclosed between a component’s start and end tags.
In this example, chunk d, containing the <img> tag, is the entire body of the
DirectLink component. The page itself has a body, the top-level static chunks
(chunks b and f) and the components that aren’t enclosed by other components
56
CHAPTER 2
Figure 2.5 The Home page template is broken into chunks of static HTML and component tags.
Static HTML chunks render as themselves; the DirectLink renders in code, in its
renderComponent() method, and causes its body (the <img> tag) to render by invoking its
renderBody() method.
(chunk
c). When the page renders, it renders just the chunks in its body. Static
HTML chunks render as themselves; they are passed on through to the client web
browser unchanged. Components are responsible for rendering themselves and
their body.
Figure 2.5 references two methods related to the DirectLink component: renderComponent() and renderBody(). The renderComponent() method is implemented
by components that render in Java code (rather than using their own template).
The method is invoked by the component’s container, in this case the Home
page itself, as part of the Home page’s render.
The second method, renderBody(), is inherited by the DirectLink component
from the AbstractComponent base class. The component invokes this method
from its own renderComponent() method to render the text and components in
its own body—the static <img> tag enclosed by the DirectLink’s <a> and </a> tags.
In this case, the body of the DirectLink is simple, static HTML. That’s often
not the case; a component may contain a mix of static HTML and other components. Tapestry figures it all out, properly slotting each chunk of the page’s template into the body of the correct component. Rendering a page is a recursive
process, since components may themselves have their own templates, containing
other components. Chapters 6 and 8 go into great detail about creating new
components, including components that have their own template.
Tapestry’s HTML template parser is very forgiving; although the examples in
this book all follow Extensible HTML (XHTML) conventions, the template parser
can handle the kind of HTML you’ll find in the wild: unquoted attribute values,
57
mixed uppercase and lowercase, single or double quotes, unquoted attribute values, and lots of additional whitespace. As elsewhere in Tapestry, if the parser is
unable to parse a template it will throw an exception providing line-precise
reporting of the problem.
The last piece of the Home page puzzle is the page class; this is where we put
our application-specific logic—the code that will actually start a new game.
2.2.3 Defining the Home page class
So, what happens when the user clicks the link that was created when the page
rendered? In Tapestry, that’s the million-dollar question,3 the point where all
this talk of simplicity, consistency, and components starts to make a difference.
Here’s the short answer: You tell the component about a method in your page
class to execute, and it executes the method when the link is clicked. Now, let’s
see what this looks like in practice. We’ll start with listing 2.5, the source code for
the Home page class.
Listing 2.5 Home.java: Java class for the Home page
package hangman1;
import org.apache.tapestry.IRequestCycle;
import org.apache.tapestry.html.BasePage;
public class Home extends BasePage
{
public void start(IRequestCycle cycle)
{
Visit visit = (Visit)getVisit();
visit.startGame(cycle);
}
}
A page class has many responsibilities defined by the framework, including the
ability to act as a container of other components. Fortunately, the BasePage class,
from which the Home class extends, contains the code needed to fulfill all these
responsibilities; for the Home page, all we need to add is the little bit of applicationspecific logic to be executed when the Start link is clicked. That logic shows up as
a method, start(), implemented by the Home page class.
3
Since Tapestry is open source, money is not the best way to gauge status. Perhaps this should be the
“million download question” instead!
58
CHAPTER 2
The start() method is a listener method, a method that will be invoked in
response to a user clicking a particular link. Its implementation is to defer to the
Visit object to actually start a new game—we’ll discuss what the Visit object is
shortly; for the moment, we’ll concentrate on how it is that the start() method
is invoked when a user clicks the link.
Listener methods are ordinary instance methods, implemented by the page’s
class, that have a specific method signature:
public void method(IRequestCycle cycle)
The method must always be public, return void, and take a single parameter of
type IRequestCycle.
Tapestry components may have any number of parameters, both optional
and required. The DirectLink component has several optional parameters and
one that’s required (listener). The binding for the listener parameter was provided in the Home page’s HTML template:
<a href="#" jwcid="@DirectLink"
listener="ognl:listeners.start">
. . .
</a>
TIP
Tapestry checks that there’s a binding for each required parameter. If
you remove the listener attribute from the HTML template for the
Home page, the page will not display. Instead, you’ll get an exception
report with this message: Required parameter listener of component Home/
$DirectLink is not bound. Home/$DirectLink is the name of the page and
the ID of the component.
The DirectLink component’s listener parameter is used to find the listener
method it should execute when the end user clicks the link visible in his or her
web browser. The ognl: prefix on the attribute value informs Tapestry that the
value is an Object Graph Navigation Language (OGNL) expression to be evaluated, rather than a literal string constant. In Tapestry terminology, the expression listeners.start is bound to the DirectLink’s listener parameter.
WARNING
Don’t forget the ognl: prefix. If you omit the prefix, Tapestry treats the
value as a string literal. Removing the prefix from the DirectLink’s listener parameter will result in an error like this when you click the link:
Parameter listener (listeners.start) is an instance of java.lang.String, which does
not implement interface org.apache.tapestry.IActionListener. When you see
59
exceptions such as this, or perhaps ClassCastExceptions within your
own code, the likely cause is a missing ognl: prefix.
How does the OGNL expression listeners.start end up executing this method?
All pages and components inherit a property, listeners, from the AbstractComponent base class. The listeners property contains a nested property for each listener method implemented by the class. Underneath the covers, there’s an
interface, IActionListener, and a little bit of Java reflection used to connect the
DirectLink component with the page’s listener method; this is shown in figure 2.6.
A class may have any number of listener methods, each with a unique and
individual name. Listener methods inherited from superclasses are also available
through the listeners property.
WARNING
If your OGNL expression references a listener method that doesn’t exist,
you’ll get an exception when you click the link. For example, changing
the expression to ognl:listeners.star results in an exception with
this message: Unable to resolve expression ‘listeners.star’ for hangman1.Home@
19b808a[Home]. You’ll also see an ognl.NoSuchPropertyException for
the property star.
An invalid listener method will result in the same exception: This will
occur if the method is not public or has the wrong method signature.
Sit back and think about this for a moment: We’ve just extended the behavior of
this page within the application by writing a very short method, the start() listener method. The provisions we’ve made in the HTML template to get this
Figure 2.6 The Tapestry servlet receives and interprets the incoming request and invokes
trigger() on the DirectLink component. The DirectLink invokes the listener method provided by
the page. After the method is invoked, a page is rendered, forming the HTML response sent back to
the client web browser.
60
CHAPTER 2
listener method to execute on cue are so minor that they’re barely worth considering. The Hangman application’s Home page is unusual in that it has just the
single bit of behavior—but you can imagine a more complicated page with many
links (and, as you’ll see in chapter 3, forms); adding each new bit of behavior is
still just…adding another listener method.
This gets to the heart of the Tapestry goals described in chapter 1:
■
Simplicity—Adding new operations takes minimal code and minimal
changes to the HTML template.
■
Consistency—Add as few or as many operations as you like, and the process stays the same. Look at any page in the application, and it still looks
the same.
■
Feedback—By working with the framework, errors in Java code, in the
template, or in the specification are detected and verbosely reported by
the framework.
A good practice is to keep listener methods short and focused on simply interfacing Tapestry components with business logic stored in domain objects. That’s
demonstrated here by having the start() listener method simply find the Visit
object and let it do the work of actually starting a new game.
2.2.4 Examining the Visit object
The Visit object is an application-wide space for storing application logic and
data. This object is accessible from all pages and components within the application and contains information specific to a single client of the web application. A single Visit object instance is shared by all pages within the application.
The object fulfills much the same role as the HttpSession does in a typical servlet application, and in fact, the Visit object is ultimately stored as an HttpSession attribute.
All web applications eventually store some form of client-specific server-side
state. The HttpSession acts like a map, storing named attributes. Simple as this
seems, in real applications, a considerable amount of code must be written to
retrieve attribute values from the HttpSession, cast them to the right type, create
them on the fly as needed, and delete them when they are no longer needed.
Here again, Tapestry steps in to rethink this approach in terms of objects,
methods, and properties. In chapter 7, we’ll cover how Tapestry allows page
properties to be stored persistently between requests, which is appropriate for
values that are used only within a single page.
61
For more general data, used throughout an application, Tapestry allows for a
single Visit object. Tapestry doesn’t know or care about the type of the Visit
object. There is no specific Visit class defined by the framework; each application defines its own Visit class. The accessor method for the Visit object provided by the page (defined by the interface IPage and implemented by the class
BasePage) doesn’t specify the type of the object:
public Object getVisit();
It then becomes a matter of casting to the application-specific type:
The Visit object is automatically created by the framework the first time it is referenced; you must configure Tapestry, providing the name of the class to instantiate (this may be configured inside the web deployment descriptor; see section 2.6).
Once the Visit object is created, it is stored in the HttpSession for persistent
access in later requests.
Developer code never has to worry about the HttpSession. The HttpSession
itself is created only as needed. A stateless application is more efficient than a
stateful one, and a Tapestry application will operate in a stateless mode until
there is actual server-side state to store. The framework takes care of this transition automatically, which would be very cumbersome to accomplish in ordinary
servlet code because each and every servlet would need custom logic to check for
the existence of the session and create it only as needed.
For our Hangman application, the Visit object is responsible for controlling
page flow. It acts as a façade around the WordSource and Game objects, handles the
process of starting a new game, and processes guesses made by the player. The
Visit class for the Hangman application is provided in listing 2.6.
Listing 2.6 Visit.java: controller object for the Hangman application
package hangman1;
public class Visit
{
private WordSource _wordSource = new WordSource();
private Game _game = new Game();
public void startGame(IRequestCycle cycle)
{
_game.start(_wordSource.nextWord());
b
Invoked by
Home page
listener method
62
CHAPTER 2
b
cycle.activate("Guess");
}
public void makeGuess(IRequestCycle cycle, char ch)
{
if (_game.makeGuess(ch))
return;
c
cycle.activate(_game.isWin() ? "Win" : "Lose");
Invoked by
Guess page
listener
method
}
public Game getGame()
{
return _game;
}
d
Provides
game
property
}
b
c
d
The startGame() method is invoked by a listener method on the Home page
to start a new game. It is also invoked by listener methods on the Win and
Lose pages.
The makeGuess() method is invoked by a listener method on the Guess page; the
listener method passes in the character to be guessed and the request cycle (so
that the Visit object can activate the Win or Lose page, if necessary).
The Game object is exposed as a read-only property of the Visit object. You’ll see
references to the properties of the Game object in the template as ognl:visit.
game.property.
When the Home page invokes the startGame() method on Visit, Visit gets a
random word and sets up the Game instance with it by invoking Game’s start()
method. The call to activate() is used to change the active application page;
the active page is responsible for rendering the response. Initially, the Home
page is the active page, because it contains the DirectLink component that was
triggered. Invoking the activate() method allows the correct page, the Guess
page, to render the response.
2.3 Implementing the Home page using
standard servlets
Despite the fact that the previous discussion about the DirectLink component,
listener methods, and the Visit object was unavoidably long-winded, in the end
we’ve shown that creating a link and getting an application-specific method to
execute when the link is clicked is extremely simple.
Implementing the Home page using standard servlets
63
Let’s see what would be involved in accomplishing the same thing using standard servlets and JSPs. In this simple example, the JSP is very straightforward—so
much so that it could as easily be an entirely static HTML page. The DirectLink
component is replaced by a standard HTML link to a servlet we’ll provide:
<a href="startGame"> <img . . . /> </a>
Of course, this example is not representative. Most application operations will
involve quite a bit more: more servlets to implement the operation, more query
parameters to fill in the details, and more code to build and interpret the
URLs—all things that the Tapestry framework provides you for free.
Regardless, this example uses a very simple operation with no parameters.
We still need to add a few lines to the application’s web deployment descriptor, web.xml:
<servlet>
<servlet-name>startGame</servlet-name>
<servlet-class>StartGameServlet</servlet-class>
</servlet>
<servlet-mapping>
<servlet-name>startGame</servlet-name>
<url-pattern>/startGame</url-pattern>
</servlet-mapping>
Finally, we need the actual servlet, shown in listing 2.7.
Listing 2.7 StartGameServlet.java: hypothetical servlet for starting a game
import java.io.IOException;
import
import
import
import
import
javax.servlet.ServletException;
javax.servlet.http.HttpServlet;
javax.servlet.http.HttpServletRequest;
javax.servlet.http.HttpServletResponse;
javax.servlet.http.HttpSession;
public class StartGameServlet extends HttpServlet
{
protected void doGet(HttpServletRequest request,
HttpServletResponse response)
throws ServletException, IOException
{
HttpSession session = request.getSession(true);
Visit visit = new Visit();
session.setAttribute("visit", visit);
c
b
Stores Visit
for later
Gets or creates
the session
64
CHAPTER 2
visit.startGame();
d
Chooses random
word to guess
RequestDispatcher d =
request.getRequestDispatcher("/Guess.jsp");
d.forward(request, response);
e
Forwards to
the Guess.jsp
page
}
}
b
c
d
e
This accesses an existing HttpSession for the client or creates a new one if necessary.
We store the Visit object as a session attribute so that it can be accessed in later
requests or by a JSP.
As in the Tapestry Hangman application, the hypothetical servlet Visit object is
responsible for selecting a random word to guess.
The RequestDispatcher object is used to bridge from the servlet to a JSP that can
render the response.
This servlet creates a Visit instance, similar in scope and implementation to
the Visit object used in the Tapestry application. Once created, the Visit
object is stored in the HttpSession, where it will be available in subsequent
requests. The implementation of this Visit class may use the same Game and
WordSource domain objects used by the real Tapestry application.
Extending this comparison from one single interaction to the innumerable
interactions in a typical web application underscores the amount of developer
effort needlessly wasted on most web applications. You are forced to drop out
of the world of objects and methods and deal directly with aspects of the
HTTP protocol and the Servlet API . You must define a URL to trigger your
operation, create a new servlet class to perform that operation, and record the
mapping from the URL to the servlet in the web.xml deployment descriptor.
In a team environment, you will be competing with your fellow developers to
update the deployment descriptor and to lay claim to the possible URLs for
the application.
Certainly, as you become more experienced writing servlet-based applications, you will find shortcuts to help you streamline this effort. Unfortunately,
different developers are quite likely to create their own suite of shortcuts. In a
large team effort, getting the bits and pieces of the application written by different developers interoperating properly can become quite a challenge
because of the impedance caused by all of the developers’ individual schemes.
When using Tapestry, this is rarely an issue because Tapestry defines a standard
Developing the Guess page
65
way for different parts of the application to interoperate—using objects, methods, and properties.
Now that we’ve seen how the Home page and the Hangman application’s
Visit object work together to start a new game, we can continue to the Guess
page, the primary page in the Hangman application.
2.4 Developing the Guess page
The Guess page is the central page for the Hangman application; it allows the
player to guess at letters of the target word. Figure 2.1 shows an example of the
Guess page in action.
The page has a number of responsibilities:
■
It displays the number of guesses remaining (as a number) as well as
the number of incorrect guesses so far (as the growing stick figure).
■
It displays the partially guessed target word, with lines replacing the as-yet
unguessed letters.
■
It displays a grid of remaining letters to guess; each letter is a clickable link.
■
It supports the “hand-scrawled” look and feel, using custom images to display numbers and letters.
To accomplish all these tasks, we’ll be introducing several new concepts for Tapestry specifications, HTML templates, and Java classes, as well as new Tapestry
components. We’ll start with the full listings for the HTML template, the page
specification, and the page class, and then show how the different responsibilities we’ve listed are implemented—as Tapestry markup in the HTML template
combined with entries in the page specification and code in the Java class. We’ll
begin with listing 2.8, the HTML template for the Guess page.
Listing 2.8 Guess.html: HTML template for the Guess page
<html>
<head>
</head>
<body>
<table>
<tr>
<td><img alt="Tapestry Hangman" src="images/tapestry-hangman.png"
</td>
66
CHAPTER 2
<td width="70" align="right"><img jwcid="@Image"
alt="ognl:visit.game.incorrectGuessesLeft"
image='ognl:getAsset("digit" +
visit.game.incorrectGuessesLeft)'
height="36" src="images/Chalkboard_3x8.png" width="36"
border="0"/>
</td>
<td><img alt="Guesses Left" src="images/guesses-left.png"
</td>
</tr>
<tr>
<td>
</td>
<td>
</td>
<td><img jwcid="@Image"
image='ognl:getAsset("scaffold" +
alt="[Scaffold]" src="images/scaffold.png" border="0"/>
</td>
</tr>
</table>
<br>
<table>
<tr valign="center">
<td width="160">
<p align="right"><img alt="Current Guess"
src="images/guess.png" align="middle" width="127" height="20"
border="0"/></p>
</td>
<td><span jwcid="@Foreach" source="ognl:visit.game.letters"
value="ognl:letter"><img jwcid="@Image"
image="ognl:letterImage" alt="ognl:letterLabel" height="36"
src="images/Chalkboard_5x3.png" width="36"
border="0"/></span>
b
c
<span jwcid="$remove$">

<img height="36" alt="A" src="images/Chalkboard_1x1.png"
width="36" border="0"/><img height="36" alt="_"
border="0"/><img height="36" alt="_"
d
e
border="0"/>
</span>
</td>
</tr>
<tr>
<td valign="top">
<p align="right"><img alt="Choose" src="images/choose.png"
height="20" width="151" border="0"/></p>
</td>
<td width="330"><span jwcid="selectLoop"><a href="#"
jwcid="select" class="select-letter"><img jwcid="@Image"
image="ognl:guessImage" alt="ognl:guessLabel" height="36"
border="0"/></a></span>
67
e
f
<span jwcid="$remove$">

<a class="select-letter" href="#"><img height="36" alt="B"
src="images/Chalkboard_1x2.png" width="36" border="0"/></a>
<a class="select-letter" href="#"><img height="36" alt="C"
<a class="select-letter" href="#"><img height="36" alt="D"
<img height="36" alt="-" src="images/letter-spacer.png"
width="36" border="0"/>
<a class="select-letter" href="#"><img height="36" alt="F"
<a class="select-letter" href="#"><img height="36" alt="G"
<a class="select-letter" href="#"><img height="36" alt="H"
<a class="select-letter" href="#"><img height="36" alt="I"
<a class="select-letter" href="#"><img height="36" alt="J"
<a class="select-letter" href="#"><img height="36" alt="K"
<a class="select-letter" href="#"><img height="36" alt="L"
<a class="select-letter" href="#"><img height="36" alt="M"
<a class="select-letter" href="#"><img height="36" alt="N"
<a class="select-letter" href="#"><img height="36" alt="O"
<a class="select-letter" href="#"><img height="36" alt="P"
<a class="select-letter" href="#"><img height="36" alt="Q"
g
68
CHAPTER 2
<a class="select-letter" href="#"><img height="36" alt="R"
<a class="select-letter" href="#"><img height="36" alt="S"
<a class="select-letter" href="#"><img height="36" alt="T"
<a class="select-letter" href="#"><img height="36" alt="U"
<a class="select-letter" href="#"><img height="36" alt="V"
<a class="select-letter" href="#"><img height="36" alt="W"
<a class="select-letter" href="#"><img height="36" alt="X"
<img height="36" alt="-" src="images/letter-spacer.png"
<a class="select-letter" href="#"><img height="36" alt="Z"
</span>
</td>
</tr>
</table>
</body>
</html>
b
c
d
e
f
g
g
This Image component selects and displays the correct image identifying the
number of incorrect guesses remaining to the player.
The second Image component selects and displays an image for the man on the
scaffold, showing how many incorrect guesses the player has made so far.
These components display the target word, with underscores marking unguessed
letters within the word.
This portion of the template is marked for removal (using the special $remove$
value for the jwcid attribute). The <img> tags within the <span> exist for
WYSIWYG preview but must be removed because they conflict with the dynamic
content provided by d.
These components provide an array of clickable letters, allowing the player to
guess the next letter in the target word.
This portion of the template is also marked for removal.
This page was converted directly from the HTML mockup; the bulk of the template consists of placeholder values (for the number of guesses, for the stick figure, for the partially guessed word, and for the grid of guessable letters) that will
actually be discarded in favor of dynamically generated HTML. We’ll go into
more detail on each portion of the HTML template shortly.
69
Listing 2.9 is the page specification for the Guess page.
Listing 2.9 Guess.page: specification for the Guess page
<page-specification class="hangman1.Guess">
<component id="selectLoop" type="Foreach">
<binding name="source" expression="visit.game.guessedLetters"/>
<binding name="value" expression="letterGuessed"/>
<binding name="index" expression="guessIndex"/>
</component>
b
<component id="select" type="DirectLink">
<binding name="listener" expression="listeners.makeGuess"/>
<binding name="parameters" expression="letterForGuessIndex"/>
<binding name="disabled" expression="letterGuessed"/>
</component>
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
name="digit0"
name="digit1"
name="digit2"
name="digit3"
name="digit4"
name="digit5"
path="images/Chalkboard_1x7.png"/>
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
name="scaffold5"
name="scaffold4"
name="scaffold3"
name="scaffold2"
name="scaffold1"
name="scaffold0"
path="images/scaffold.png"/>
path="images/scaffold-1.png"/>
c
d
<context-asset name="space" path="images/letter-spacer.png"/>
<context-asset name="dash" path="images/Chalkboard_5x3.png"/>
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
name="a"
name="b"
name="c"
name="d"
name="e"
name="f"
name="g"
name="h"
name="i"
name="j"
name="k"
e
70
CHAPTER 2
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
name="l"
name="m"
name="n"
name="o"
name="p"
name="q"
name="r"
name="s"
name="t"
name="u"
name="v"
name="w"
name="x"
name="y"
name="z"
e
</page-specification>
b
c
d
e
The <component> element is used to declare components. The type of component and the configuration of the component’s parameters go here, in the page
specification.
The <context-asset> element defines an asset file that is stored within the web
application context. This first set of assets includes the digits used to display the
number of remaining incorrect guesses. These assets are given logical names
that are referenced in the Java page class.
The second group of <context-asset> elements defines the images used for the
stick figure.
The remaining <context-asset> elements define the images used for the letters of
the alphabet, as well as a blank space image and the underscore image (as dash).
This is a much longer specification than for the Home page, and it demonstrates
a couple of new features: the ability to define the type and configuration of components in the specification rather than in the HTML template, and the ability to
define assets, which are named references to static files such as images or
stylesheets. Again, we’ll revisit the relevant portions of this specification shortly.
Finally, listing 2.10 is the source for the Guess class: the Java class for the
Guess page.
Listing 2.10 Guess.java: Java class for the Guess page
package hangman1;
import org.apache.tapestry.IAsset;
public class Guess extends BasePage
{
private char _letter;
private boolean _letterGuessed;
private int _guessIndex;
public void initialize()
{
_letter = 0;
_letterGuessed = false;
_guessIndex = 0;
}
b
Resets page
properties
public char getLetter()
{
return _letter;
}
public void setLetter(char letter)
{
_letter = letter;
}
public String getLetterLabel()
{
return ("" + _letter).toUpperCase();
}
public IAsset getLetterImage()
{
if (_letter == '_')
return getAsset("dash");
c
Converts letter
property to an
image
return getAsset("" + _letter);
}
public boolean isLetterGuessed()
{
return _letterGuessed;
}
public int getGuessIndex()
{
return _guessIndex;
}
public void setLetterGuessed(boolean letterGuessed)
{
_letterGuessed = letterGuessed;
}
71
72
CHAPTER 2
public void setGuessIndex(int guessIndex)
{
_guessIndex = guessIndex;
}
public IAsset getGuessImage()
{
if (_letterGuessed)
return getAsset("space");
d
String name = "" + getLetterForGuessIndex();
Converts
guessIndex
property to
an image
return getAsset(name);
}
public char getLetterForGuessIndex()
{
return (char) ('a' + _guessIndex);
}
public String getGuessLabel()
{
if (_letterGuessed)
return " ";
char ch = Character.toUpperCase(getLetterForGuessIndex());
return new Character(ch).toString();
}
public void makeGuess(IRequestCycle cycle)
{
Object[] parameters = cycle.getServiceParameters();
Character guess = (Character) parameters[0];
e
char ch = guess.charValue();
Visit visit = (Visit) getVisit();
Listener
method
invoked when
a letter is
clicked
visit.makeGuess(cycle, ch);
}
}
b
c
d
This method is invoked when the page is created and at the end of each request,
to reset any properties back to pristine values, ready for the next request.
This method creates a read-only, synthetic property, letterImage, that provides
the correct image for whatever the letter property currently is.
Likewise, this guessImage property returns the correct image based on the guessIndex and letterGuessed properties.
e
73
This listener method is invoked when a letter image is clicked; it exists to determine the correct parameters to pass to the Visit object’s makeGuess() method.
Guess is a typical Tapestry page class; it contains properties and methods that
support the rendering of the page as well as listener methods activated by links
on the page.
2.4.1 Displaying the remaining guesses
The first dynamic bit is the part of the HTML template that displays the number
of incorrect guesses remaining to the player:
<img jwcid="@Image"
height="36"
src="images/Chalkboard_3x8.png"
This snippet has an array of responsibilities:
■
It must render an HTML <img> tag and fill in a number of attributes
dynamically.
■
It must convert the incorrectGuessesLeft property of the Game object into
a string, as the alt attribute.
■
It must select the correct image file to display the number of guesses left
and build a URL to that file (as the src attribute).
Earlier we saw how the DirectLink component on the Home page inserted an
<a> tag into the response sent to the client web browser. The Image component,
another standard Tapestry component, is actually much simpler; it inserts an
<img> tag, generating the tag’s src attribute from its image parameter. Here we
want it to provide the correct image (one of the hand-drawn digits) and the corresponding alt value.
NOTE
To support WYSIWYG editing, the HTML template uses an <img> tag,
knowing that the component will, at runtime, render an <img> tag. The
Image component will override the src attribute in the template, which
is also here just to help with the WYSIWYG preview of the template.
In a Tapestry template, each component must have properly balanced start and
end tags. An alternative, used here, is to include an XML -style empty tag, one
74
CHAPTER 2
that ends with />. Tapestry is flexible about attribute quoting; because the image
parameter’s expression uses double quotes, the entire expression is enclosed in
single quotes.
WARNING
Match your open and close tags. You must supply a matching close tag
for each component’s start tag. Tapestry even checks that all the start
tags and end tags on a page properly nest (it is forgiving for all tags that
aren’t components). Changing the end of the <img> tag from /> to just
> will result in the following exception: Closing tag </td> on line 13 is improperly nested with tag <img> on line 12. Tapestry matched the </td> on
line 13 with the <td> on line 12 (before the <img> tag) and realized that
the <img> tag hadn’t yet been closed, even though it’s a component.
The first OGNL expression, visit.game.incorrectGuessesLeft, is very straightforward; it retrieves the incorrectGuessesLeft property from the Game object
(via the Visit object). The incorrectGuessesLeft property (a number) is converted to a string and becomes the value for the <img> tag’s alt attribute. In the
client web browser, this value becomes the tooltip for the image and is also used
for accessibility (visually impaired users may have the value read to them by
their computer).
The other expression, for selecting the image is more complicated. It also
obtains the incorrectGuessesLeft property, but then it uses that value as a parameter
when invoking the getAsset() method on the page. This underscores why OGNL
is so useful and powerful; without OGNL, this access and manipulation would
have to occur in Java code. Using OGNL, we are able to assemble the complete
string and invoke a Java method, getAsset(), on our page, all in one step. The
invoked method returns the asset object representing the image to use, which is
ultimately converted into a URL by the Image component and inserted in the
HTML response as the src attribute of the <img> tag.
NOTE
4
Using OGNL expressions where possible allows you to assume a rapid
application development cycle, free from the normal edit/compile/deploy cycle that occurs with Java code. You can simply edit your templates and specifications in place to see changes.4 Later, you can recode
It is possible to disable the normal caching that occurs inside Tapestry so that templates and specifications are reread for each new request. This allows changes to templates and specifications to take
effect immediately. Consult the Tapestry reference documentation, distributed with the framework,
for the details.
75
OGNL expressions as Java methods for greater application efficiency.
Another good option, when not prototyping, is to move nontrivial
OGNL expressions into the page specification (an example of this is
shown in section 2.4.3); this results in a much improved separation of
the View from the Model and application logic, which ultimately yields a
more maintainable application.
Tapestry allows you, as the developer, to decide how pure a separation between the View and the Model you will maintain. At one extreme,
the pragmatic view, you may put as much logic (in the form of OGNL expressions) as you want directly into the HTML template. This pushes together purely presentation-oriented aspects of the application (such as
layout and fonts) with the behavioral aspects of the application (shown
in this example as references to page properties, including visit and
visit.game). Such an approach is perfectly acceptable for prototypes,
or for small projects where a strong separation between developers isn’t
realistic. Most of the examples in this book use this pragmatic approach
simply because it puts related information side by side, making it easier
to comprehend.
At the other extreme, the purist view, your HTML template contains
only placeholders for components; all details about the component configuration are stored outside the template, in the page specification.
This is critical on larger projects, where a division can be expected between the HTML developers responsible for page mockups and the
Java developers responsible for converting the mockups into a working
application. Minimizing how much of the application’s implementation
is exposed to the HTML developers reduces the potential for conflicts
between the Java developers and the HTML developers.
Assets are any kind of file that may be distributed as part of the WAR; the most common types of assets are images and stylesheets. The Image component’s image
parameter expects an asset object (an object that implements the IAsset interface),
not a string, and this pairs up with the getAsset() method, which returns just such
an object. The getAsset() method is inherited from the AbstractComponent base
class; it allows access to the named assets defined in the page specification.
The names of the assets come from the <context-asset> elements in the page
specification (in listing 2.9). What’s happening is a mapping from a logical name
(such as x or dash) to a particular file (such as images/Chalkboard_4x6.png or
images/Chalkboard_5x3.png). The assets abstraction has some other important
uses related to localization and to packaging components into reusable libraries.
Those uses are covered in more detail in chapters 6 and 7.
76
CHAPTER 2
Defining assets in the page specification
The page specification for the Guess page declares assets for the letters, digits,
and underscore as well as all the images of the stick figure on the gallows. The
Guess page specification includes the following lines to declare the six digits
used in the user interface:
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
WARNING
name="digit0"
name="digit1"
name="digit2"
name="digit3"
name="digit4"
name="digit5"
Tapestry checks that a file matching the provided asset path exists.5
This check occurs when the page specification is first read and takes
place regardless of whether anything ever uses the asset. Putting a typo
into one of the names in the previous snippet results in the following
exception: Unable to locate asset ‘digit0’ of component Guess as context:/images/Challkboard_1x7.png.
Here, we can see how the aliasing is useful. The letters and numbers were initially drawn onto a grid, and a slicing tool was used to generate a set of individual files from the cells of the grid. The filenames provided by the slicing tool are
not intuitive (they are based on the position in the grid, rather the value of the
image, and so are somewhat arbitrary), but the use of assets allows the code to
reference them using more friendly names. Of course, we could have simply
renamed the files output by the slicing tool, but by leaving the names as is, we
can change the original letter grid image and then use the same slicing tool to
regenerate all the images without having to go through the painful renaming
process a second time. Tapestry has provided a little bit of abstraction and flexibility that ultimately makes the build process for this application more agile,
because an annoying manual step (renaming the files) is not necessary.
Assets also provide a separation of concerns, dividing the HTML developers
from the Java developers. For example, an HTML developer may decide to redo
the graphics for the page and use a new tool to generate the images of the digits—which would result in new filenames, possibly even new types (perhaps GIF
or JPEG), but no change to the logical names of the assets. Either the HTML
5
This applies to the context assets defined here and the private assets we’ll discuss in chapter 6. A third
asset type, the external asset, is not checked.
77
developer or the Java developer would need to update the page specification to
change the filenames, but there would be no change to the HTML template or
even to the Java class (if the Java class ever accessed any assets by name).
Now that we have a way of mapping from logical names to actual asset files,
we still need a way to figure out which logical name, and thus, which asset,
should be used when displaying the remaining guesses.
Calculating the right asset
Displaying the digit image is a matter of selecting the correct asset as the image
parameter to the Image component. This occurs in the HTML template using an
OGNL expression:
image='ognl:getAsset("digit" + visit.game.incorrectGuessesLeft)'
Here, OGNL has done something fairly complex: building up the name of the
asset and invoking the page’s getAsset() method. There are penalties, however:
This chunk of text is somewhat unwieldy and forces us to use single quotes, since
the expression itself contains double quotes. Putting OGNL expressions into your
template, especially expressions of this complexity, is not much better than putting Java scriptlets into a JSP: Such OGNL expressions strongly tie together the
presentation of the page with the implementation.
One option would be to move more of this logic into equivalent Java code.
This can be easily accomplished by referencing a new, read-only property in the
HTML template:
<IMG jwcid="@Image"
image="ognl:digitImage"
Reference to the page’s
height="36"
digitImage property
We would then implement an accessor method for this new digitImage property
in the Guess class:6
public IAsset getDigitImage()
{
int guessesLeft = visit.getGame().getIncorrectGuessesLeft();
return getAsset("digit" + guessesLeft);
}
6
Because this approach is only hypothetical, you won’t see this method in the Guess class in listing 2.10.
78
CHAPTER 2
Another option, which we’ll explore shortly, is to move the OGNL expression into
the page specification. The decision to use OGNL expressions, Java code, or
some mix of the two is left to you, according to your personal taste and the particular situation. The modest runtime performance penalty for using OGNL is
easily offset by increased developer productivity.
Using informal component parameters
If you check the description for the Image component in appendix C, you’ll see
that it defines two possible parameters: a required image parameter and an
optional border parameter. However, if you run the application and view the
source of the page, you’ll see that the other attributes included in the <img> tag
in the template (alt, width, and height) are still present in the <img> tag rendered by the Image component. How can this be?
The majority of Tapestry components, including Image and DirectLink, allow
informal parameters. Informal parameters are additional parameters for the component beyond those that are formally declared by the component. These additional parameters are simply added to the rendered tag as additional attributes.
Informal parameters can be unevaluated static values, such as for width, or
expressions, such as for alt. Some informal parameters are discarded so that
they don’t conflict with attributes rendered directly by the component. For example, it doesn’t matter that the template provides a value for the src attribute (in
the <img> tag for the Image component); the value in the template is discarded
because the Image component will itself generate an src attribute from the asset
provided in the image parameter. The src value in the template exists to support
WYSIWYG previewing of the template; its value is discarded in favor of the real,
dynamic URL computed on the fly in the live application. Only components that
map directly to an HTML tag will accept informal parameters; each component
indicates within its own component specification whether it accepts or discards
informal parameters.
So, when the Image component renders, it will mix and match the informal
parameters with the HTML attributes it generates from formal parameters. This
is a capability missing from JSP tags, where specifying an undeclared JSP tag
attribute is simply an error. With JSP tags, you are limited to just the attributes
explicitly declared for the tag, no more.
Displaying the right stick figure image
Continuing with the rest of the Guess page, the next dynamic section of the
HTML template is also related to the incorrectGuessesLeft property; it is used
79
to display one of several images for the gallows, showing increasing amounts of
the stick figure as the incorrectGuessesLeft property drops toward zero.
<img jwcid="@Image"
image='ognl:getImage("scaffold" +
alt="[Scaffold]"
src="images/scaffold.png"
border="0"/>
Again, we use the same trick; we come up with a logical name for the image
asset and map that logical name to an actual file by way of the <context-asset>
elements in the page specification. This is a good, simple example of the MVC
pattern in action; the Model in this case is the Game object and its incorrectGuessesLeft property, but there are two Views of the data: the first as a digit, the
second as the stick figure on the gallows.
The remaining dynamic portions of the page are more complex and require
using multiple components in concert to produce the desired output.
2.4.2 Generating the guessed word display
The next section of the Guess page displays the target word the player is
attempting to guess, or at least as much of the target word as the player has
guessed so far. Generating this portion of the page starts with the Game object,
which has a property, letters, for just this purpose. The letters property is an
array of each letter of the target word as an individual character. Each unguessed
letter in the target word is replaced with an underscore character.
As with the previous examples, we can’t simply output the individual letters as
characters. To keep the hand-scrawled look and feel, each letter must be translated to the correct image. The template uses two different components to generate the display: a Foreach component (which performs a kind of loop)
enclosing another Image component. The two components work together to display one letter after another.
<span jwcid="@Foreach"
source="ognl:visit.game.letters"
value="ognl:letter">
<img jwcid="@Image"
image="ognl:letterImage"
alt="ognl:letterLabel"
height="36"
width="36"
border="0"/>
</span>
80
CHAPTER 2
Looping with the Foreach component
Foreach is a looping component; it iterates over the list of values provided by its
source parameter 7 and updates its value parameter for each value from the
source before rendering its body. This is a crucial feature of Tapestry component
parameters; by binding a property to a component parameter, the component is
free not only to read the value of the bound property, but also to update the
property as well.
The Foreach component is represented in the template using a <span> tag,
which is very natural: The HTML <span> tag is simply a container of other text
and elements in a page. It doesn’t normally display anything itself, but is commonly used in conjunction with a stylesheet to control how a portion of a page
is rendered.
Although the Foreach’s location in the template is specified using a <span>
tag, when it renders, it does not produce any HTML directly; it simply renders
the text and components in its body repeatedly. The sequence is shown in
figure 2.7.
So, the Foreach component will render its body many times, but that doesn’t
help the Image component display the correct letter image. Just before the
Foreach renders its body (on each pass through the loop), it sets a property of the
page to the next letter in the word (from the array of characters provided by the
Game object). The trick is to convert this letter into the correct image. The Guess
page class includes a property, letter, which is bound to the Foreach component’s value parameter so that it can be updated by the Foreach:
private char _letter;
public char getLetter()
{
return _letter;
}
public void setLetter(char letter)
{
_letter = letter;
}
7
The Foreach component is flexible about how it defines “a list of values.” It may be an array of objects,
or a java.util.List, or even a single object (which is treated like an array of one object).
81
Figure 2.7 The Foreach component reads a list of values bound to its source
parameter from a domain object (which is often the page that contains the
component). For each item in the list, it updates a domain object property bound
to its value parameter, and then renders its body. Components within its body
can get the value from the domain object property.
NOTE
In chapter 3, we’ll see how Tapestry can automatically create properties
at runtime (and the benefits of doing so beyond less typing). For now,
we’ll mechanically code these properties ourselves by supplying the instance variable and pair of accessor methods.
82
CHAPTER 2
Translating letters to images
Once again, we are using assets to obtain the correct image to display within the
page. The assets for the letters a through z are named, simply, a through z. However, there’s a gotcha for the underscore character; its asset name is dash.
The Guess page class implements another method to provide the asset to display:
{
if (_letter == '_')
return getAsset("" + _letter);
}
This simple method captures the special rule about replacing the underscore
character with the asset named dash. The Foreach component is responsible for
invoking setLetter() with the correct letter well before getLetterImage() is
invoked by the Image component.
NOTE
Because this method is public and follows the naming convention for a
JavaBeans property, it can be referenced in the HTML template as
ognl:letterImage. This is a common approach in Tapestry—creating
synthetic properties, properties that are computed on the fly, rather than
just exposing a value in an instance variable.
The letters in the list (provided by the Game object) are all lowercase, but the tooltip (generated from the <img> tag’s alt attribute) looks better if the letter is
uppercase. This is another, minor example of the Controller (the page) mediating between the Model (the Game object) and the View (the Image component
within the HTML template). This case conversion is accomplished by binding the
value for the alt parameter to the letterLabel property of the page. The
getLetterLabel() accessor method simply converts the letter to uppercase and
returns it as a string:
{
char upper = Character.toUpperCase(_letter);
return new Character(upper).toString();
}
Removing unwanted portions of the template
If you examine the complete HTML template in listing 2.4, you’ll see that just
after the </span> tag for the Foreach component is a long chunk of additional
83
images—images for additional letters from the target word, as dashes. These
images were copied over from the original HTML mockup and are left in place so
that the HTML template will still preview properly. Without these additional
images, the target word will appear as a single underscore, which may not be
enough to validate the layout of the page. At the same time, these extra images
must not be included in a live, rendered page or the target word will appear to
be six letters longer than it actually is.
Earlier, you saw that Tapestry will drop unwanted HTML attributes that are
provided in HTML tags to support WYSIWYG preview. This is a larger case, where
an entire section of HTML is dropped. The block to be removed is surrounded by
a <span> tag:
<span jwcid="$remove$"> . . . </span>
The special component ID, "$remove$", is the trigger for Tapestry’s template
parser that this portion of the HTML template should be discarded. This is a
second aspect of instrumenting an HTML mockup into an HTML template:
marking portions of the mockup for removal, yet leaving them in for previewing purposes.
So far on this page, we’ve covered just output-only behaviors: displaying
the right digit image, or the right letter from the target word. The most
involved part of the page comes next—the part that allows players to select
letters to guess.
2.4.3 Selecting guesses
This portion of the page is a grid of letters that the player may click on to make
guesses. As usual, the letters are represented as images, to keep with the handscrawled look and feel. As the player makes guesses, the guessed letter is erased,
and one or more positions in the target word are filled in or another segment is
added to the stick figure.
To accomplish this, we’ll use a combination of components: another Foreach
to iterate over the different letters of the alphabet, a DirectLink to create a
link, and an Image to display either the image for the letter or the image for a
blank space for an already guessed letter. The three components appear in the
HTML template:
<span jwcid="selectLoop">
<a href="#"
jwcid="select"
class="select-letter">
<img jwcid="@Image"
84
CHAPTER 2
image="ognl:guessImage"
alt="ognl:guessLabel"
height="36"
width="36"
border="0"/>
</a>
</span>
Two of these components look a little sparse compared to previous examples;
that’s because we’ve chosen to use the declared component option for them
rather than configure them in-place as implicit components. For a declared component, we just put the component ID in the HTML template. Tapestry recognizes that the value for the first jwcid attribute is just an ID and not a
component type, because it does not contain the @ character (as the previous
usages of components have done). For a declared component, the element in the
HTML template is simply a placeholder; the jwcid attribute provides a component ID that is used to link to a <component> element in the page specification.
The type and configuration of the component is provided in the page specification itself:8
<component id="selectLoop" type="Foreach">
<binding name="source" expression="visit.game.guessedLetters"/>
<binding name="value" expression="letterGuessed"/>
<binding name="index" expression="guessIndex"/>
</component>
<component id="select" type="DirectLink">
<binding name="listener" expression="listeners.makeGuess"/>
</component>
The information that goes into the specification is the same as what would be put
directly into the HTML template, but the format is slightly different. In the
HTML template, we must mark OGNL expressions with the ognl: prefix; but in
the XML we have a specific element, <binding>, that is always an OGNL expression (other elements are used for literal strings and other variations). There is no
difference to Tapestry whether a component is declared in the specification or in
8
The template may still specify additional formal and informal parameters. In keeping with the goal
to provide the clearest separation of presentation and logic, the informal parameters, which are most
often related purely to presentation, should go in the template, and the formal parameters, which are
most often related to the behavior of the component, should go in the page specification.
85
the HTML template; here, the sheer number of parameters for the two components indicated that specification was a better home for the component configuration than the HTML template.
WARNING
Mistakenly using the ognl: prefix inside a page or component specification will create an OGNL expression that is invalid. You’ll see an exception,
such as Unable to parse expression ‘ognl:visit.game.guessedLetters’. The fact that
the ognl: prefix shows up in the exception message as part of the expression is the indicator that you included the prefix where it is not allowed.
Once again we are combining the behaviors of different components and using
the page to mediate between them. We are also making use of new features of
the Foreach and DirectLink components by binding additional parameters of
the components.
Getting the images for the letters
The source of all this data is the guessedLetters property of the Game object; this
is an array of 26 flags, one for each letter in the alphabet. Initially, all the flags
are false, but as the player makes guesses, the corresponding flags are set to true.
The Foreach component will loop through the 26 flags and set the letterGuessed property of the page to true or false on each pass through the loop. In
addition, binding the index parameter of the Foreach component directs it to set
the guessIndex property of the page. This value starts at zero and increments
with each pass through the loop. The other components simply translate from
this ordinal value to a letter in the range of a to z. This functionality is implemented by additional properties and methods in the Game class, as shown in the
following snippet:
private boolean _letterGuessed;
private int _guessIndex;
public boolean isLetterGuessed()
{
return _letterGuessed;
}
public void setLetterGuessed(boolean letterGuessed)
{
_letterGuessed = letterGuessed;
}
public int getGuessIndex()
86
CHAPTER 2
{
return _guessIndex;
}
public void setGuessIndex(int guessIndex)
{
_guessIndex = guessIndex;
}
public char getLetterForGuessIndex()
{
return (char) ('a' + _guessIndex);
}
Getting the right letter image for the current letter within the loop is very similar
to the previous examples. Although the dash will never occur, we do have to substitute a blank image for any letter that has already been guessed:
public IAsset getGuessImage()
{
if (_letterGuessed)
String name = "" + getLetterForGuessIndex()
return getAsset(name);
}
That covers how we get the image for each letter display, but what about the link
that the player uses to make a guess?
Handling the links for guesses
Were we to display the link for guesses using ordinary servlets, we’d define a
query parameter whose value is the letter selected; that is, we would encode the
letter into the URL. Since we’re using Tapestry, we don’t want to think in terms of
query parameters, but instead, we want to think of objects and properties—but
we still want the URL to carry this piece of information. When we render the link,
we know which letter the link is for, and when the link is clicked, we need that
information back. In Tapestry terms, we need to invoke a specific listener
method (as before on the Home page) but also propagate along some additional
data: the letter selected by the player.
We’ll use a DirectLink component, as we did with the link on the Home page,
but with two differences. First, we only want to display the link itself (the <a> and
</a> tags) some of the time; we want to omit the link for letters that have already
been guessed (the positions that show up as blank space), because letters may
87
only be guessed a single time. Second, we need a way to know which letter has
been selected. The DirectLink component includes formal parameters to satisfy
both of these needs.
The disabled parameter is used to control whether the link renders the <a>
and </a> tags. The disabled parameter is optional, and by default, the link is
enabled. A DirectLink component will always render its body, regardless of the
setting of the disabled parameter. The Guess page binds the disabled parameter to the letterGuessed property of the page—the same property that is set by
the Foreach component and used in the getGuessImage() method:
This ensures that once a letter has been guessed, there will not be another link
for that letter. Shortly, we’ll see how we also ensure that the guessed letter is
replaced by a blank space. Once again, we are working at the level of objects and
properties, and not treating all of this HTML rendering as just a text processing
problem. A common, ugly “JSP-ism” is to use embedded scriptlets to avoid writing the open and close tags, wrapping the <a> and </a> tags inside conditional
blocks, which can be a messy affair.
The DirectLink embodies the Tapestry philosophy, solving a similar problem
using JavaBeans properties and Tapestry component parameters. Every component decides, in its own code, whether to render; the DirectLink has a small conditional statement to control whether an <a> element is rendered—but that’s
Java code in a Java file, not cluttering up a JSP. The end result is a cleaner, simpler, easier-to-use solution.
To identify which letter is actually clicked by the player, we will use yet
another component parameter, named parameters. We can bind a single value,
or an array, or a java.util.List to the parameters parameter, which, like the
disabled parameter, is optional (we didn’t use it before with the Start link on the
Home page). The collection of values provided by the parameters parameter is
recorded into the URL constructed when the DirectLink component renders.
When the link is submitted, the array of parameters is reconstructed and is available to the listener method.
For this case, we use a single value, provided by the property letterForGuessIndex:
Each time the DirectLink component renders, within the Foreach component
loop, the value for the letterForGuessIndex property will reflect the current letter
88
CHAPTER 2
in the loop and the URL written into the HTML response will be different, as a
portion of the URL will be an encoding of the letterForGuessIndex property.
When the link is clicked, the listener method can get the parameters back:
public void makeGuess(IRequestCycle cycle)
{
Object[] parameters = cycle.getServiceParmeters();
Character guess = (Character) parameters [0];
char ch = guess.charValue();
visit.makeGuess(cycle, ch);
}
The parameters encoded into the URL by the DirectLink are available in the listener method as an array of object instances, which can be obtained from the
getServiceParameters() method of the IRequestCycle object. Even when, as in
this case, there’s only a single parameter value, an array is returned. The lone
character value is the first and only element in the array.
In addition, the value has been converted from a scalar type, char, to a wrapper object type, Character, but it is a simple chore to convert it back. The parameter value is not simply converted to a string; it retains its original type (which is
encoded into the URL along with the value). You can see a bit of this in the web
browser’s Address field in figure 2.1; the URL shown contains much information
used by Tapestry, but at the end is cp, an encoding of character p (the player had
just clicked the letter P). Chapter 7 discusses how Tapestry encodes information
into URLs.
From here, it’s simply a matter of obtaining the Visit object and letting it do
the rest of the processing of the player’s guess, which may result in a win or a loss
or more guessing. Because we pass the request cycle to the Visit, this object is
fully capable of selecting which page will render the response by invoking the
activate() method on the request cycle.
Adding this new interaction, the handling of guesses by the player, involved
little more than creating the new listener method and pointing the DirectLink
component at the method. Without Tapestry, this same functionality would
entail not only writing a servlet and registering it into the web deployment
descriptor, but also creating code to generate the hyperlink in the first place.
This latter code could take the form of Java scriptlets in the JSP, or a new JSP
tag in a JSP tag library. In either case, the HTML in the JSP file would deviate
Configuring the web.xml deployment descriptor
89
further from ordinary HTML, and the ability to preview the web page would be
diminished. With Tapestry, the HTML template will continue to look and act
like standard HTML.
Instead, we are making use of existing components, the DirectLink, and a
consistent approach to encoding data into the URL. Once again, we’re seeing the
consistency goal: Anywhere in the application where we have a link that needs to
pass along some data in the URL , we can use and reuse the same tool, the
DirectLink component and its parameters parameter. In addition, because Tapestry properly encodes the data type with the data (rather than just converting
all the parameters to strings), we can consistently pass any type of data in the
URL: strings, characters, numbers, or even custom objects.
That wraps up the Guess page; we’ve discussed how to extract information
from the Game domain object and present it in various ways and also figured out
how to react to user input. We’ve kept the domain logic (in the Game and WordSource objects) separate from the presentation logic (the Guess page class,
HTML template, and page specification), using listener methods and the Visit
object as the bridge between the two aspects.
2.5 Developing the Win and Lose pages
The other two pages in the application, Win and Lose, are displayed when the
player successfully guesses the word, or when the player exhausts all his or her
incorrect guesses. There is nothing new on these pages; they duplicate bits and
pieces of the Home and Guess pages. In fact, there’s a bit of unwanted duplication in the HTML templates: the Java code and the page specifications. In
chapter 6 we’ll see how easy it is to create new components that encapsulate
this functionality and remove this duplication. Remember: More code means
more bugs!
Our Hangman application is nearly complete; all that’s left is to fulfill our
contract with the servlet container and create a deployment descriptor for the
Hangman application WAR.
2.6 Configuring the web.xml deployment descriptor
All of these HTML templates and page specifications do not automatically
become a web application. We still need a servlet to act as the bridge between the
Servlet API and the Tapestry framework. Fortunately, this does not require any
coding, since the framework includes the necessary servlet class. All that’s neces-
90
CHAPTER 2
sary is to configure the web deployment descriptor, which is the file WEB-INF/
web.xml. The deployment descriptor is provided in listing 2.11.
Listing 2.11 web.xml: web deployment descriptor for the Hangman application
<!DOCTYPE web-app
PUBLIC "-//Sun Microsystems, Inc.//DTD Web Application 2.2//EN"
"http://java.sun.com/j2ee/dtds/web-app_2_2.dtd">
<web-app>
<servlet>
<servlet-name>hangman</servlet-name>
<servlet-class>org.apache.tapestry.ApplicationServlet
</servlet-class>
<init-param>
<param-name>org.apache.tapestry.visit-class</param-name>
<param-value>hangman1.Visit</param-value>
</init-param>
<load-on-startup>1</load-on-startup>
</servlet>
<servlet-mapping>
<servlet-name>hangman</servlet-name>
<url-pattern>/app</url-pattern>
</servlet-mapping>
</web-app>
The deployment descriptor maps an instance of the Tapestry ApplicationServlet
to the path /app within the servlet context. Using /app as the servlet path is a
common convention for Tapestry applications but not a requirement; Tapestry
will adapt to whatever servlet mapping is actually used.
The servlet context’s name is based on the name of the web application
archive (the WAR file), which is hangman1.war; therefore, the complete URL of
the servlet is http://localhost:8080/hangman1/app. This means that the base URL
for the application is http://localhost:8080/hangman1/, which is why relative
URLs (in static HTML) to assets such as images/guess.png work, both when previewing the HTML template and at runtime.
The org.apache.tapestry.visit-class initial parameter is used to tell Tapestry what class to instantiate as the Visit object.
Using the <load-on-startup> element in the deployment descriptor is recommended, especially during development. Loading on startup causes the application servlet to be instantiated and initialized. Often, problems in the application or
Summary
91
deployment descriptor will be detected immediately at startup; this is an even better idea for more advanced applications that use an application specification.9
2.7 Summary
In this chapter, we’ve seen the basics of creating a web application using Tapestry. A Tapestry application is divided into individual pages; those pages are constructed by combining components, an overall HTML template, and a small
amount of Java code. Tapestry leverages the Model-View-Controller pattern to
isolate domain logic from the user interface. We’ve also begun to see the “light
touch” of Tapestry, where simple properties and short Java methods are woven
together to create very complex, dynamic, interactive user interfaces.
This simple application demonstrates some of the key patterns that occur
when developing in Tapestry. It shows how components interact with each other
by reading and setting properties. It shows how the page can act as a Controller,
coordinating the domain logic and mediating between its embedded components. We’ve also demonstrated how easy it is to add new interactions to a page,
in the form of listener methods.
We’ve begun to demonstrate how Tapestry, by excusing developers from mundane “plumbing” tasks, really frees up developer energies. It enables you to
implement more complicated behaviors in much less time and be more confident that your code is bug free. Tapestry can give projects the one thing money
truly can’t buy: time—time to test and debug back-end code, time to locate and
fix performance problems, even time to add new features.
9
Application specifications are an optional file described in chapter 6. They are needed only to access
some advanced feature of Tapestry, such as referencing a component library.
Tapestry and
HTML forms
This chapter covers
■
How HTML forms work
■
Creating simple forms using Tapestry
components
■
Creating page properties using property
specifications
■
Creating forms containing loops
92
93
Handling links and images is a good start, but the lifeblood of any real web application is the HTML form. Individual links, as in the Hangman example from chapter 2, provide a limited form of interactivity: Users will click a link, or they won’t.
HTML forms provide a rich form of expression for the user—the ability to enter
specific values in fields, whether by typing in a text field, clicking a check box, or
selecting an option from a drop-down list.
Tapestry’s approach to handling HTML forms follows its general philosophy
of hiding the details of HTTP and the Servlet API and operating in terms of
objects, methods, and properties. In Tapestry, a group of components work
together when a form is submitted to update page properties before invoking a
listener method.
In this chapter, we’ll cover components for creating basic HTML form elements, including text fields, radio buttons, and check boxes. In chapter 4, we’ll
examine more advanced form-related components that include additional client- and server-side support. Chapter 5 discusses Tapestry components devoted
to validating user input.
The source for the examples for this chapter and the following chapters is
available online (see appendix B). Once you deploy the samples into Tomcat,
you can access them by opening a web browser to http://localhost:8080/examples/
app. You will find a listing of the examples, organized by chapter.
3.1 Understanding HTML forms
HTML forms allow us to gather complex information from users of our web
applications. The use of forms is ubiquitous on the Web. Without support for
forms, the Web would never have exploded; limited to just hyperlinks, it would
likely have stayed as passive, static hypertext—an interesting toy, but with no
possibility of creating an Amazon or an eBay. Without forms, there would be no
way to build any of the interesting e-commerce, content management, or community sites.
HTML forms are significantly different from forms within a desktop application. In a desktop application, each component (text field, check box, push button) interacts with the application completely independently of the others. The
user may update one field without changing the value for another field. The
code for a desktop application is informed of every key press or mouse click, a
luxury not possible with HTML forms. HTML forms are displayed to the user,
who may use the mouse or the Tab key to move between the form controls and
enter values. When the user clicks the submit button, the values for all of the
94
CHAPTER 3
fields and controls within the form are packaged together in a single request to
the server. The server sees only the final result, not the individual mouse movements and key presses, or even the individual changes to fields.
Within an HTML form, various HTML elements work together. The outermost element, <form>, groups together the many other elements it encloses
between its start and end tags. Other elements within <form> are used to create
types of form controls; these are listed in table 3.1.
Table 3.1
Elements used to create form controls
HTML element
Control type
<input type="checkbox">
Check box (toggles on or off)
<input type="radio">
Radio selection (user selects one value from a list of values)
<input type="text">
Simple text input field
<input type="password">
Password field (text field where user input is obscured)
<input type="hidden">
Hidden field (not visible to the user)
<input type="submit">
Form submit button; sends the request to the server
<input type="reset">
Reset button; returns all fields to initial values
<input type="image">
Image map field; submits form and identifies where, within the
image, the user clicked
<input type="file">
File upload
<select>
Drop-down list or multiple-selection list
<textarea>
Multiline text input field
Each of these form control elements has a name attribute. When the form is submitted, the value for each control is submitted as a query parameter using the
control’s name. Figure 3.1 shows how this comes together; when the page containing the form is rendered, a name is assigned to each element. Once the form
is submitted, these values are provided in the form submission and available
through the Servlet API’s HttpServletRequest object. Figure 3.1 shows a single
form; beginning at the top center, the form starts as HTML: <form> and <input>
tags. The client web browser uses these to construct form controls (radio buttons,
text fields, and so forth) for the user to directly interact with. When the user finishes entering information and clicks the submit button, the request includes the
values for those fields as query parameter values. Within the Servlet API, those
query parameters are accessible via the HttpServletRequest object.
95
Figure 3.1 The form and the data it contains are represented in different ways at different times.
The form starts as HTML sent via HTTP to the web browser. The web browser displays controls the
user can interact with. The form submission contains the values entered by the user, which finally
show up as Java objects.
In a servlet application, the next step is to access those query parameters and use
them to update domain object properties. (This may involve some conversions,
such as parsing string values into numbers.) One of the stumbling blocks when
you’re using servlets is ensuring that the form control names used in the JavaServer Page (JSP) match the query parameter names referenced in the servlet
code. With servlets, you are responsible for coming up with appropriate names
and keeping the JSP and the servlet synchronized with respect to those names.
Expecting form control names in the JSP to stay synchronized with Java code in a
servlet is an example of weak binding (discussed in chapter 1) between the JSP
and the servlet.
Tapestry adds a layer beyond the query parameters—a layer of components.
The form control components read page properties when rendering the initial
HTML. The same components come back into play when the form is submitted
to read the query parameters and update the page properties.
The framework takes care of all the mundane plumbing: providing names for
form control elements, writing HTML, and reading query parameters. There is
never a danger of the HTML template for a form getting out of synchronization
96
CHAPTER 3
with the code that handles the form submission because that logic is distributed
into the very components that render the form’s HTML in the first place.
All of this means that by the time your form’s listener method is invoked, the
components will have already updated all of your domain objects’ properties to
match the values entered by the user. The listener method can simply use those
properties to accomplish the operation the form is intended for, whatever that
may be.
To appreciate how all these concepts fit together, we’ll start with a simple example before getting into the nitty-gritty of the individual Tapestry components.
3.2 Creating a simple login form
One of the most common, and basic, uses of HTML forms is to authenticate a
user: collecting a username and password to identify the specific user to the
application. An e-commerce application will need to know who, specifically, the
user is before processing an order; most other categories of applications have a
similar need to know the identity of the user. We’ll start our discussion of Tapestry HTML form support with a simple version of such a form, shown in figure 3.2.
The HTML template for this page is provided in listing 3.1.
Figure 3.2 The Login page after the user has entered an incorrect username
and password.
Creating a simple login form
97
Listing 3.1 Login.html: HTML template for the Login page
<html jwcid="@Shell" title="Login">
<body>
b
<span jwcid="@Conditional" condition="ognl:message">
<font color="red">
<span jwcid="@Insert" value="ognl:message">
Error Message
</span>
</font>
</span>
c
<p/>
<form jwcid="@Form" listener="ognl:listeners.login">
<table>
<tr>
<th>User Name:</th>
<td>
<input type="text" jwcid="@TextField"
value="ognl:userName"
size="30"/></td>
</tr>
<tr>
<th>Password:</th>
<td>
<input type="password" jwcid="@TextField"
value="ognl:password"
hidden="ognl:true"
size="30"/>
</td>
</tr>
<tr>
<td><input type="submit" value="Login"/></td>
</tr>
</table>
d
e
f
</form>
</body>
</html>
b
The Shell component is a standard Tapestry component for rendering the
<html>, <head>, and <title> elements. It has a number of other uses, as you’ll see
later, such as simplifying references to Cascading Style Sheets (CSS).
98
CHAPTER 3
c
d
e
f
If there is an error, such as an unrecognized password, then the page will be
redisplayed with a non-null value in its message property. This block of the template is used to format and display that message.
A Form component, not a simple <form> element, must be used, and must
enclose the other form control components. The component will render the URL
for the form submission, in much the same way as a DirectLink component. The
provided listener will be invoked when the form is submitted.
The TextField component will read the userName property when the page renders. When the form is submitted, it will update the same property.
This TextField component is configured, via its hidden parameter, as a password
field. Tapestry uses one component, TextField, for both ordinary and password
text fields.
Like the DirectLink component used in the Hangman application in chapter 2, a
Form component has a listener parameter bound to an application-provided
listener method. This particular Form component encloses some static HTML
and two form control components. Both of these components are TextFields.
The first TextField component edits the userName property of the page. The second TextField component edits the password property of the page and also renders as a password field (where the characters typed by the user are not visible).
As in previous examples, additional attributes are provided so that the element
will preview correctly in an HTML editor.
In most web application frameworks (including Jakarta Struts), you are
responsible for orchestrating things so that the names of HTML form elements
(and thus the names of query parameters supplied when the form is submitted)
match the names of domain object properties. This kind of “engineered coincidence” creates unwanted dependencies between the text fields and the domain
objects and can open up other limitations. (For example, what if there is no specific property? What if different fields update different objects?) Tapestry avoids
these dependencies; the component ID, the name of the query parameter, and
the Object Graph Navigation Language (OGNL) expression used to read and
update the property will all be different, and the developer will be concerned
only with the OGNL expression.
We’ve also introduced two new components that are not related to forms. The
Shell component is a handy shortcut for outputting the <html>, <head>, and
<title> elements for a page. The Conditional component evaluates a condition
(an OGNL expression) and either renders its body (the portion of the page template enclosed by its start and end tags) or skips over its body. Here, the Conditional component is used to display an error message if the message is not null.
99
Conditional is very flexible as to what types are acceptable for its condition
parameter. Here the parameter is bound to a string property, and the Conditional treats null or the empty string as false and a non-empty string as true.
As with the Hangman application pages in chapter 2, understanding the
Login page involves more than just the HTML template; the application-specific
code for the form submission is in the Java class.
3.2.1 Implementing the Login page class
Listing 3.2 shows the Java side of the form: the implementation of the Login
page class and, specifically, the listener method invoked by the Form component.
Listing 3.2 Login.java: Java class for the Login page
package examples;
public abstract class Login extends BasePage
{
public abstract String getUserName();
public abstract String getPassword();
public abstract void setMessage(String message);
b
c
public void login(IRequestCycle cycle)
{
if (isValidLogin(getUserName(), getPassword()))
{
cycle.activate("Main");
return;
}
d
e
setMessage("Invalid user name or password.");
f
}
private boolean isValidLogin(String userName, String password)
{
return "tapestry".equalsIgnoreCase(userName);
}
g
}
b
These methods are abstract because Tapestry will create the actual properties at
runtime (it will create a subclass with the necessary methods and fields). This is
controlled by the page’s specification.
100
CHAPTER 3
c
d
e
f
g
The listener method is invoked when the form is submitted, after the properties
have been updated.
The abstract methods are used to access the information submitted with the form.
On success, the Main page is activated; a real application would do a lot more
bookkeeping here.
On error, the message property is updated. The active page is not changed, so
the Login page will be redisplayed (showing the error message, and the username and password entered by the user).
This is just a placeholder for the work a real application would perform; presumably, the code would use information in a database to authenticate that the username is known and that the password is valid.
By the time the Form’s listener method has been invoked, the properties of the
page (userName and password) will already have been updated with values provided by the user and submitted in the request as query parameters. The listener
method is not responsible for extracting query parameters or any of the other
plumbing normally associated with processing a form submission. In this simplified example, a valid login will result in sending the user to the Main page of the
application (not shown in this example). An invalid login will present the Login
page again, with a visible error message.
Here you see the simplicity and consistency goals pop up again; form submissions result in updates to properties of the page, followed by the invocation of
the listener method for the Form component. There’s no question about query
parameter names—the information submitted in the form just shows up as properties that are accessible to the listener method. Any question of accessing the
information submitted in the form becomes simple access to page properties set
by the form control components. All the HTML forms in your application will be
variations on this same basic pattern. Forms will always operate consistently
because there’s just one approach, and the normal stumbling blocks (such as
what to name form controls and query parameters) just aren’t your concern;
they’re entirely under the control of the Tapestry framework.
An additional look at listing 3.2 shows one oddity: The class itself is abstract
and has abstract methods for accessing the userName, message, and password
properties. These methods are abstract because we are letting Tapestry create the
properties dynamically at runtime as specified properties.
101
3.2.2 Using specified properties
Creating JavaBean properties for a typical JavaBean object is normally straightforward. A property consists of a private instance variable and a pair of accessor
methods (one to read the value, one to update the value):
private Type _myProperty;
public Type getMyProperty()
{
return _myProperty;
}
public void setMyProperty(Type newValue)
{
_myProperty = newValue;
}
The naming of the methods is defined by the conventions of the JavaBeans framework. In fact, the JavaBeans framework works backward from the method names
(setMyProperty() and getMyProperty()) to deduce a property named myProperty.
This is the essence of the contract between the JavaBean and the JavaBeans framework—a simple, loose, flexible contract based on this naming convention.
Using Tapestry, it is possible to create new properties on pages and components automatically. The page (or component) specification will define the property
name and type, and Tapestry will create the property: the instance variable and the
pair of accessor methods. At runtime, Tapestry actually creates a subclass of your
class, which fills in these details and instantiates an instance of the subclass.
Creating new properties in this way is useful for three reasons:
■
Creating properties manually is a boring, mechanical process.
■
Tapestry ensures that property values are reset at the end of the request to
default values automatically (this step is necessary to properly support
reuse of the page object in later requests—a topic we discuss in more detail
in chapter 7).
■
Properties can be defined as persistent, meaning that they are stored on the
server between request cycles. This approach to managing server-side state
is introduced later in this chapter and also discussed in detail in chapter 7.
In your code, such as the Login page code in listing 3.2, you will define abstract
methods to read or update the properties. You have to define only the methods
you specifically need—the Tapestry-fabricated subclass will include getter and
setter methods for each property you specified in the page specification. You
102
CHAPTER 3
never have to provide the implementations of these methods; Tapestry creates
those implementations for you.
NOTE
This approach is not unique to Tapestry. When using entity Enterprise
JavaBeans (EJBs) using container-managed persistence (CMP) 2.0, you
code in the same way: creating and invoking abstract accessor methods,
with the expectation that the application server will provide the necessary implementations in a subclass.
The creation of new properties is triggered in the page (or component) specification.
We provide the specification for the Login page in listing 3.3. Each <propertyspecification> element defines a property that we want to add to the page.
Listing 3.3 Login.page: specification for the Login page
<page-specification class="examples.Login">
<property-specification name="message" type="java.lang.String"/>
<property-specification name="userName" type="java.lang.String"/>
<property-specification name="password" type="java.lang.String"/>
In this case, all three properties are strings, but properties of all types (including
primitive Java types) may be created in this way.
This Login page example is just a starting point for the kinds of forms supported by Tapestry. Tapestry has a variety of components that correspond to all
the types of HTML form elements, shown in table 3.2.
Table 3.2
HTML form elements and corresponding Tapestry components
HTML element
Tapestry component
Discussed in…
<form>
Form
Chapter 3
<input type="text"/>
TextField or ValidField
Chapters 3, 5
<textarea>
TextArea
Chapter 3
<input type="password"/>
TextField
Chapter 3
Table 3.2
103
HTML form elements and corresponding Tapestry components (continued)
HTML element
Tapestry component
Discussed in…
<select>
Select or PropertySelection
Chapter 3, 4
<option>
Option
Chapter 3
<input type="radio"/>
Radio (and RadioGroup)
Chapter 3
<input type="checkbox"/>
Checkbox
Chapter 3
<input type="submit"/>
Submit
Chapter 3
<input type="image"/>
ImageSubmit
Chapter 3
<input type="hidden"/>
Hidden, ListEdit
Chapter 4
<input type="file"/>
Upload
Chapter 4
You’ve now seen the very basics of Tapestry’s approach to HTML forms. Form
control components can read properties when a page is rendered and then
update the same properties when the user submits the form. To understand
more, let’s start with a detailed look at the Form component and how it relates to
the other components.
3.3 Understanding the Form component
Central to Tapestry’s approach to HTML forms support is the Form component.
It is not enough to simply use a <form> tag in a Tapestry page; all the form control components (for creating text fields, check boxes, and such) require that
they be enclosed by a Form component. The Form component’s responsibilities
start with the rendering of the <form> tag, but beyond that, the Form component
orchestrates all the form control components enclosed by the Form component,
both during the initial render and when the form is submitted. The Form component is combined with the other form-related components listed in table 3.2.
To appreciate what the Form component does and why it is so necessary, you
may find it useful to look at the steps you would take when creating an HTML
form (and a servlet to process the form submission) without using Tapestry.
3.3.1 Developing forms without Tapestry
Creating functioning HTML forms in an ordinary servlet application shifts most
of the responsibility onto your shoulders. You start by identifying the purpose of
the form: what business process the form will initiate, or what properties of
which domain objects the form will update (or some combination of the two).
104
CHAPTER 3
You then select which types of form controls will be used: text fields, check boxes,
radio buttons, and so forth. Each of these controls is represented by a particular
HTML element and configuration.
The next step is to select a unique name for each form control, typically a
name with some meaning, such as productId, password, or userName. Most often,
the name selected represents how the value will be utilized when the form is submitted; it may be the name of a property of an object, or the name of a parameter to some operation. From here, the last two steps are to create the JSP
containing the <form> and the form control elements, and to create a servlet to
process the form submission.
The end goal of the form submission is to execute some form of business process. Much of the time, this business process is designed to create or update data
stored in a database. Other examples of business processes are more involved,
such as submitting an order within an e-commerce application. In any case,
these business processes are hungry for data from the user; that data arrives in
the form query parameters and must be extracted from the request, converted,
and validated before it can be used.
The servlet will need to know which names have been assigned to which form
controls, because those names become the names of the query parameters, accessible via the HttpServletRequest object provided to the servlet. The servlet must
extract the value for each query parameter, and then possibly convert the value
from a string to an appropriate type (say, an integer or a date) and assign the
value to the correct domain object property or business process parameter.
There may also be some validations on individual fields or on the form as a
whole. In some cases, the validations may direct that the form be redisplayed
with an error message so that the user may make corrections before resubmitting
the form. Validations may be on individual fields, such as checking that a
numeric field contains a numeric value. Other validations may be based on comparing the values for multiple fields (for example, a form that allows a user to
input a date range will check that the start date precedes the end date). Still other
validations may have some external dependency; for instance, a servlet that processes an application login form must validate that the username and password
are correct and will redisplay the page with an error message if they are not.
Assuming all validations pass, the next step for the servlet is to initiate a business process using all the collected parameters, and the final step is to send a
response page back to the client.
Using Tapestry, nearly all of this work is eliminated.
105
3.3.2 Developing forms with Tapestry
Returning to the world of Tapestry, the framework’s goal is once more to offload
as much of this work as possible onto the framework. It is still the responsibility
of the developer to handle the first stages: to identify the business process and
decide which form control elements will be used. From there, the developer
begins to leverage Tapestry to do the grunt work. Anything involving HTTP
query parameters falls into this category.
Tapestry has a number of components that correspond to each of the standard HTML form control elements. Table 3.2 is a simple mapping between the
HTML form control elements and the Tapestry form control components.
The framework supports two kinds of forms: simple and complex. The Login
page is a simple form; it always renders with the same structure—a <form> element enclosing two text fields. In the real world, forms are not always so simple;
a common example is a “shopping cart” page in an e-commerce application that
shows multiple line items. It’s entirely reasonable in such a form for the user to
change quantities for several items within an order before submitting it. The
exact set of elements in such a form is determined at runtime and will be different for different users at different times. Behind the scenes, there is some form
of Order object and a collection of LineItem objects. Rendering such a page
(using a JSP or a Tapestry HTML page) involves some form of loop over the available LineItems.
The Form component takes on the responsibility for assigning unique IDs for
the form control elements. These IDs are used as the name attribute in the HTML
element and become the name of the query parameter when the form is submitted. Having the Form component assign these IDs might seem unnecessary; after
all, couldn’t the components just use their component ID? In simple forms, that
is exactly what happens, but not all forms are simple. What happens in a complex form, one where some components render multiple times within a loop?
Each time a particular component renders, it must have a unique element ID so
that the component can get a single value from a query parameter when the
form is submitted. At the same time, it is desirable that complex forms with loops
and conditionals be handled in the same way as simple forms. To support simplicity and consistency, this decision about what element ID to use for each component is handled by the Tapestry framework, not the developer.
As each form control component renders, it locates the Form component that
encloses it. The component obtains its element ID from the Form. This element
ID will be used for the name attribute of the element, which is later used as the
106
CHAPTER 3
name of the query parameter for the element when the form is submitted. The
Form ensures that the names it provides to the components are unique. Generally, the name will match the component’s ID, but if the form element component is inside a loop (such as a Foreach component), each render of the component
will receive a different unique ID.
The form control components also read page properties as they render and
write them as the value attribute of the element. The process continues from there
through the other components enclosed by the Form. Finally, once the entire
body of the Form component has been rendered, the Form component writes the
</form> tag and the rendering of the rest of the page continues as normal.
The result of rendering the Form (from the Login example) is shown in
figure 3.3. The output HTML is a mixture of static HTML from the template and
Figure 3.3 The Form component writes several hidden control fields as well as rendering its
body. The TextField components render themselves, with names obtained from the Form.
107
dynamically generated HTML produced by the Form and TextField components.
The Form component writes the <form> and </form> elements, as well as several
hidden form fields that store control information needed by the Form to handle
the submission.
The TextField components render the <input type="text"> elements, with
the value for the name attribute supplied by the enclosing Form component. The
names are based on the IDs of the TextField components. Since no explicit ID
was assigned for either TextField component, Tapestry assigned unique component IDs for the TextFields based on their component type—"$TextField" and
"$TextField$0"—and these component IDs are then used as the form element
IDs. As you’ll see later in the chapter, a Form incorporating any kind of loop will
use even more mangled element IDs—but none of these IDs is ever visible to an
application’s end user or relevant to you as the application developer.
In figure 3.3, the two TextField components omitted the value attribute
because the corresponding page properties, userName and password, were null. If
the Login page was redisplayed because of an error (such as an invalid password), the page properties would be read and used for the value attribute.
These descriptions include a few simplifications. Tapestry forms are actually
very flexible:
■
A page may contain any number of Form components. They all work independently of each other, in just the same way that a page may contain any
number of independent DirectLink components.
■
A Form may itself be contained within another component; it doesn’t have
to be directly located within the page’s template.
■
A component enclosed by a Form may itself (in its own template) contain
form control components, such as TextField or Checkbox. This is entirely
valid. Tapestry dynamically finds all the components ultimately enclosed
by the Form during the render, working its way through the page’s template and the templates for any components within the page.
■
Form control components are not limited to editing a page property; using
OGNL expressions, they can edit properties of any object reachable from
the page.
So far, we’ve only discussed what occurs when the page containing the form renders. Eventually, the user will submit the form, and then the more interesting part
begins—moving values out of the query parameters and into page properties.
108
CHAPTER 3
3.3.3 Handling form submissions
When a form is submitted, Tapestry must perform the same kind of operations
that would occur in a non-Tapestry application:
■
Determine the names of query parameters.
■
Extract the values for query parameters.
■
Perform any necessary conversions (for example, string to integer).
■
Assign the converted values to the correct page properties.
Further, Tapestry needs to take into account any portions of the original form
that rendered inside a loop or conditional. The difference between a traditional
servlet application and a Tapestry application is that there is no single place
where this mapping between query parameters and domain object properties
takes place. The information is scattered across the many form control components enclosed by the Form. In a pure servlet application, this information will
be present in two places: inside the JSP containing the form, and within the corresponding servlets’ doPost() method. Within a Tapestry form, this mapping
information is not “stored” anywhere but is computed on the fly during the render, and those computations must be reproduced as part of the form submission
if all the data submitted in the form is to be extracted and applied to the correct
objects and properties.
Tapestry takes an unusual approach to linking query parameter names (in the
form submission) back to the correct components: It recovers all those names
and relationships by rendering the Form again, as part of the form submission
processing. This render is called the rewind phase. The rewind phase allows the
Form to walk through all the components it encloses, visiting each in the same
order as in the original render, even when the Form contains loops and conditionals. In this way, each form control component will be visited in the same
order as in the initial render, and will obtain the exact same element ID as it did
in the initial render.
During the rewind, each form control component will obtain its element ID
from the Form, just as it did in the original render. Unlike the original render,
where a component parameter is read and HTML is produced, during the rewind
phase the component will instead read a query parameter value, convert the
value as necessary, and update a domain object property.
Only the Form and its body, not the entire page, are rendered. Mixed into the
Form may be other components that are not form related—if they produce
HTML output during the rewind phase, it is simply discarded.
109
The rewind starts with the Form component itself, which renders its body.
Eventually, this render will reach a form control component, such as a TextField
or Checkbox. The component will acquire its element ID from the Form component, exactly as it did when it rendered. Each component can also determine,
from the Form, whether a rewind is taking place. Using the element ID, the component can extract the correct query parameter from the incoming request and
assign its value to a page property. The rewind then continues through any other
components also contained within the Form component.
This process is flexible and allows for cases where portions of the Form’s body
are conditional, or where a looping component, such as a Foreach, is involved.
As long as the rewind of the Form goes through the same exact steps as the original render, all the names and components will line up, the proper query parameters will be read by the matching components, and the values will be applied to
the correct page and domain object properties. It is only after the rewind phase
that the Form’s listener method is invoked.
Tapestry records into the form, as a hidden field, the list of all the element
IDs it allocates during the render process (this is shown in figure 3.3). It uses this
information when the form is submitted to determine whether the rewind
matches the render. Any deviation immediately causes an exception to be thrown
detailing the cause of the mismatch.
When such a mismatch occurs, Tapestry does not display the normal Exception page; instead the StaleLink page is displayed. A stale link indicates that the
form submission doesn’t match against the current state of the application. This
can happen when some bit of information stored on the server (possibly in a
database) changes between the time that the page containing the form is rendered and the time the form itself is submitted. Stale links are most often provoked by a Form containing a conditional or loop based on server-side state.
Included in the output on the StaleLink page is a message identifying the
point at which a mismatch occurred. On a page such as the Login form, it is
impossible for a rewind mismatch to occur, because the form is simple: There are
no looping or conditional components within the body of the Form component.
We’ll return to this subject in section 3.5, and discuss techniques for avoiding
stale links in chapter 4.
3.4 Using basic form control components
Tapestry form element components fall roughly into two categories: basic and
advanced. The basic components correspond more or less directly to HTML
110
CHAPTER 3
form elements such as <input type="text"> and <select>. Table 3.3 describes
the basic form components covered in this chapter. More advanced components
are covered in chapters 4 and 5.
Table 3.3
Basic form control components
Component
Description
Checkbox
Creates a check box that edits a boolean property bound to its selected
parameter.
Form
Creates the <form> element enclosing other components.
ImageSubmit
Creates a clickable image that submits the enclosing form.
Option
Creates an <option> element that edits a boolean property bound to the
selected parameter.
Radio and RadioGroup
Creates a number of mutually exclusive options. A single property, bound to
the RadioGroup’s selected parameter, is edited.
Select
Creates a <select> element that can enclose Option components.
Submit
Creates a button that submits the enclosing form.
TextArea
Used for accepting text input larger than a single line, corresponding to a
<textarea> form element. TextArea edits a string property bound to its
value parameter.
TextField
Used to create a simple text-entry field, either <input type="text"> or
<input type="password">. TextField edits a string property bound to its
value parameter.
Appendix D contains a complete reference to all Tapestry components.
3.4.1 Understanding the essentials
Each of the basic form control components has a similar purpose and a similar
pattern of usage. These components exist to edit a property. When a form control
component renders, it reads a property (bound to one of its component parameters) and uses that value to construct the HTML element and attributes it writes
into the response. When the form is submitted, the same component will read a
query parameter value and use that value to update the same property.
We’ve seen this already, with the TextField component, which drops quite easily into place inside an HTML template, as shown in the Login example at the
beginning of this chapter. Using the TextField component is as simple as this:
■
■
Placing the component within the HTML template, enclosed by a Form
component
Binding its value parameter to the page property to be edited
111
Each of the components is specialized to reflect the particular type of control
that will be presented to the user. This specialization is not just the HTML that
the component outputs, but even reflects the names and types of parameters
used by the component. We’ll start with the simplest component, the Checkbox,
which edits a boolean property.
3.4.2 The Checkbox component
An HTML check box is created using the <input type="checkbox"> element. A
check box control appears as a small box, which may be clicked to toggle between
a checked and an unchecked state—a check mark appears or disappears.
When the enclosing form is submitted, a query parameter for the check box is
only sent if the check box is checked. The value for the query parameter will be
the literal string "on" (unless the <input> element’s value attribute is specified,
in which case that is used as the value).
A Tapestry Checkbox component is used to edit a boolean page property.
The property we want to edit is bound to the component’s selected parameter.
The property is read when the component renders, and the component will
include the selected attribute of the <input> element if the property is true.
When the form is submitted, the Checkbox component will read the query
parameter and update the page property to either true or false, depending on
whether the check box was checked or unchecked.
For example, the following HTML template snippet is used to edit the accepted
page property:
<input type="checkbox" jwcid="@Checkbox"
selected="ognl:accepted"> I accept the terms and conditions.
3.4.3 Radio and RadioGroup components
In an HTML form, a radio element (<input type="radio">) is used to provide a
number of options from which the user can select a single option. Each radio element provides a single option, and all the radio elements in a group share a single name. When the enclosing form is submitted, the value of the query
parameter will match the value of the selected radio control.
In Tapestry terms, a single property will be edited by a group of related Radio
components. On submission, the selected Radio component will be used to figure out the correct value to assign to the edited property.
To accomplish this, Tapestry uses a second component, RadioGroup. A
RadioGroup encloses one or more Radio components. RadioGroups may not be
nested, and Radio components must be enclosed by a RadioGroup—but they
112
CHAPTER 3
don’t have to be directly enclosed by the RadioGroup; they can be spread out
about the HTML template, mixed with static HTML and other components. The
RadioGroup component identifies the property that is being edited by using its
selected parameter. Each Radio component provides a particular value, bound
to its value parameter, which will be assigned to the property if that component
is selected. The value is not limited to a simple string value; the Radio and
RadioGroup components will allow any type of object to be the value.
The RadioGroup and Radio components work together to write out the
HTML element for each Radio component. The RadioGroup provides a unique
option ID for each Radio that becomes the value attribute of the <input> element. The RadioGroup also tracks the current value of the property so that one
and only one Radio component will be rendered as the selected element.
What’s important is that the value attribute of each <input> element not
match the value parameter of the Radio component—it must be an arbitrary
option ID value instead. The value parameter may be any type: an integer, a
string, or even a custom Java object. The value attribute can only be a string, and
it’s up to Tapestry to bridge any gap between these two representations. When
the form is submitted, the RadioGroup and Radio components work backward
from the selected option ID. When a particular Radio component recognizes that
it was the option selected by the user, it reads its value parameter and provides it
to the RadioGroup, which can then update the page property bound to its
selected parameter.
For example, the following HTML template snippet is used to set the selectedSize property of the page to one of three values—5, 25, or 100:
<span jwcid="@RadioGroup" selected="ognl:selectedSize">
<input type="radio" jwcid="@Radio" value="5"/> Small
<br/>
<input type="radio" jwcid="@Radio" value="25"/> Medium
<br/>
<input type="radio" jwcid="@Radio" value="100"/> Large
</span>
When the page containing this snippet renders, the <span> for the RadioGroup
will disappear and the name attribute for each <input> element will be taken from
the RadioGroup. Assuming the initial value for the selectedSize property is 25,
the snippet will render as follows:
<input type="radio" name="$RadioGroup" value="0"/> Small
<br/>
<input type="radio" name="$RadioGroup" value="1"
113
selected="selected"/> Medium
<br/>
<input type="radio" name="$RadioGroup" value="2"/> Large
NOTE
The names for these radio elements come from the enclosing RadioGroup component. $RadioGroup (in this example) is a unique ID for
the particular RadioGroup component. In a page containing many RadioGroups, it will be a different value. One of the responsibilities of the
Form component is to provide such unique element IDs to the form
control components it encloses—but you never need to be concerned
with these.
Although we’ve managed to automatically update a property without resorting to
writing any Java code, this is still (by Tapestry standards) an unacceptably large
amount of effort to update a single property. It requires one Radio component
for each possible option, plus the RadioGroup component enclosing the other
components. In practice, most Tapestry applications make use of the PropertySelection component instead, which is described in chapter 4. The PropertySelection can, as a single component, render itself as a drop-down list or as a
table of radio buttons. In some cases, the HTML designer will want to have the
radio buttons spread out within the form (not clustered together as shown here),
interspersed with other components and static HTML—in that case, the combination of RadioGroup and Radio is appropriate.
3.4.4 Select and Option components
The HTML <select> and <option> elements are used together to create two
types of user interface controls: drop-down lists and multiple-selection lists. A
drop-down list is used to allow a single option from a list of options to be selected,
much like using radio controls. Each option is displayed as a string.
A multiple-selection list allows the user to select multiple options from a list.
The list of possible options appears in a box, which displays a limited number of
options. If there are many options, the selection box will include a scroll bar. The
user may click an option within the list or, by using a platform-specific modifier
key (the Ctrl key, under Windows), select multiple options.
Each option has a label, which is a text string visible in the list, and a value,
which is the value communicated in the query parameter if the option is selected
when the form is submitted.
These two components are used together to support single or multiple selection from a list of items. The Select component must enclose the Option components. In effect, these components work together to act as a group of Checkbox
114
CHAPTER 3
Figure 3.4
The user is provided with a multiple-selection list to choose toppings.
components. Each Option component edits an individual boolean property of
the page, indicating whether that particular Option was selected or deselected
when the form was submitted.
Internally, the Select and Option components work much like the RadioGroup and Radio components. The Select component is responsible for providing option IDs to the Option components during the render and informs the
Option components if they were selected during the rewind.
A typical use for Select and Option is to allow the user to select several items,
which are represented as a set. The following snippets show how to present a list of
items and handle multiple selections of those items. In this example, it is a list of
toppings that can be added to a hamburger. Figure 3.4 shows the example page.
At the core of this page’s HTML template are three components: a Select
component enclosing a Foreach component, which in turn encloses an
Option component:
<select jwcid="@Select" multiple="ognl:true">
source="ognl:allToppings"
value="ognl:topping">
<option jwcid="@Option"
selected="ognl:toppingSelected"
115
label="ognl:topping"/>
</span>
</select>
The Select component encloses the Foreach and Option components. Its multiple parameter is set to true, enabling multiple selection instead of the default
behavior (single selection). The Foreach component iterates through a list of
possible toppings (provided by the page’s allToppings property), updating the
page’s topping property for each one.
The Option component edits a boolean toppingSelected property, but what
we ultimately want is a java.util.Set of selected toppings. This is accomplished
by creating a synthetic property—a property that isn’t simply a wrapper around
an attribute, but is computed on the fly from other properties. The synthetic
property toppingSelected is true when the current topping (the topping property)
is in the set of selected toppings, and false otherwise. Updating the toppingSelected property adds or removes the current topping from the set. Listing 3.4
shows how easily this can be implemented.
Listing 3.4 Toppings.java: Java class for the Toppings page
package examples;
import java.util.HashSet;
import java.util.Iterator;
import java.util.Set;
import
import
import
import
org.apache.tapestry.IRequestCycle;
org.apache.tapestry.event.PageEvent;
org.apache.tapestry.event.PageRenderListener;
org.apache.tapestry.html.BasePage;
public abstract class Toppings extends BasePage
implements PageRenderListener
{
private static String[] TOPPINGS =
{ "Lettuce", "Tomato", "Cheese", "Onions", "Pickles", "Relish",
"Mustard", "Ketchup" };
public String[] getAllToppings()
{
return TOPPINGS;
}
public abstract String getTopping();
public abstract void setSelectedToppings(Set toppings);
116
CHAPTER 3
public abstract Set getSelectedToppings();
public boolean isToppingSelected()
{
return getSelectedToppings().contains(getTopping());
}
b
public void setToppingSelected(boolean toppingSelected)
{
if (toppingSelected)
getSelectedToppings().add(getTopping());
else
getSelectedToppings().remove(getTopping());
}
c
public void selectToppings(IRequestCycle cycle)
{
String toppings = getToppingsList();
ToppingsResult page =
(ToppingsResult) cycle.getPage("ToppingsResult");
page.setToppings(toppings);
cycle.activate(page);
}
private String getToppingsList()
{
if (getSelectedToppings().isEmpty())
return "No toppings.";
StringBuffer buffer = new StringBuffer();
int count = getSelectedToppings().size();
int x = 0;
Iterator i = getSelectedToppings().iterator();
while (i.hasNext())
{
if (++x > 1)
{
if (x == count)
buffer.append(" and ");
else
buffer.append(", ");
}
String topping = (String) i.next();
buffer.append(topping);
}
117
buffer.append(".");
return buffer.toString();
}
public void pageBeginRender(PageEvent event)
{
setSelectedToppings(new HashSet());
}
d
}
b
c
d
The toppingSelected property is computed by seeing if the current topping (as
set by the Foreach component) is in the selectedToppings set.
Updating the toppingSelected property adds or removes the current topping
from the selectedToppings set.
This method is defined by the PageRenderListener interface, which is called as
the page begins to render (or when a form within the page starts to rewind). It
gives us a chance to initialize the selectedToppings property to an empty
java.util.Set and thus avoid NullPointerExceptions that would occur inside
getToppingSelected() and setToppingSelected().
The Toppings page specification includes specifications for two properties:
<property-specification name="selectedToppings"
type="java.util.Set"/>
<property-specification name="topping" type="java.lang.String"/>
In addition, the class implements an interface, PageRenderListener , and a
method, pageBeginRender() . Inside the setToppingSelected() method, the
selectedToppings property is updated, but what is its initial value? The default
initial value for properties is null, which will cause a problem unless we ensure
that an instance of java.util.Set is in place in the selectedToppings property
before setToppingSelected() is invoked.
That’s what the PageRenderListener interface and the pageBeginRender() method
do: By implementing the interface, the page instance will receive a notification as the
page starts to render. This event notification takes the form of an invocation of the
pageBeginRender() method defined by the interface. The implementation for the
Toppings page creates a new empty java.util.HashSet and assigns it to the selectedToppings property. At the end of the request, the topping and selectedToppings
properties will automatically be reset to their default value: null.
Back to the toppingSelected property. This is an example of the page acting
as Controller, mediating between the Model objects and the page’s components.
118
CHAPTER 3
In this case, the Model is simply the selectedToppings set. The Foreach component will cycle through each of the known toppings, updating the topping property with a string such as "Lettuce" or "Pickles" on each pass through its loop.
The toppingSelected property is the current topping’s representation in the
selectedToppings set. When the toppingSelected property is updated, it either
adds or removes the current topping from the set (though removal is not necessary, since the set starts out empty). This all works together so that, by the time
the form’s listener is invoked, the selectedToppings set contains just the toppings that the user selected in the form.
The flexibility of the Select and Option components becomes apparent when
we go beyond selecting a number of strings to selecting complex objects. For
example, perhaps we are selecting from a list of Product objects (in an e-commerce
application) or some other kind of object pulled from a database. It may not be
possible to represent the entire object, or even its primary key, as a string. That
doesn’t affect the Select and Option components, because the value attribute
that’s recorded in the form is simply an index number, not the real value.
Option components and Checkbox components are similar from a programming model point of view. In most cases where multiple selection is not allowed,
the PropertySelection component (described in chapter 4) is a better and simpler choice. In chapter 6, we introduce a more complex component, the Palette,
which is suitable for handling multiple selections but requires that the client web
browser execute a significant amount of client-side JavaScript.
3.4.5 Submit and ImageSubmit components
The HTML <input type="submit"> form element creates a button control. Clicking the button submits the form for processing by the server. The button control’s label matches the value attribute of the <input> tag.
A form may have any number of submit controls. When the form is submitted, the control’s value attribute is the value for the query parameter; this makes
it possible to determine which submit control was used to submit the form.
The weakness of the submit control is its appearance; it is always a clickable
button with a text label. Most modern web applications demand precise control
over the look and feel of the application and need to use an image to control the
form submission. A second type of control, <input type="image">, creates an
image that the user may click to submit the form.1
1
The image control actually submits a pair of query parameters, the x and y position within the image
clicked by the user. Like many technologies on the Web, the intended purpose is rarely used.
119
Tapestry includes two components corresponding to these two controls: Submit for a standard submit button, and ImageSubmit for a clickable image. The
two components differ mostly in how they render HTML; they handle the form
submissions in the same way—using one of two patterns to distinguish which
component within the form was responsible for the submission.
The first pattern is to set a flag to indicate which component was clicked to
submit the form. The components support two parameters for this purpose:
selected and tag. The tag parameter is read, and the value retrieved is assigned
to the selected parameter if and only if the component was responsible for the
form submission. For example, a pair of buttons could be used to determine
whether the user wishes to move an item up or down within a list:
<input type="submit" jwcid="@Submit"
selected="ognl:button"
tag="up"
label="Move Up"/>
selected="ognl:button"
tag="down"
label="Move Down"/>
If the first Submit component is clicked, the button property of the page will be
set to the literal value up. If the second Submit component is clicked, the
button property will be set to the literal value down. If the form is submitted
otherwise, perhaps by pressing Enter within a text field, the button property
will not be updated.
The Form’s listener method can query the button property to decide how to
accomplish the user’s request. There is no limitation on the type of property to
be updated (the tag value doesn’t have to be a string; it can be any type of
object), and as always, the tag parameter may be an expression instead of a literal value.
The second pattern involves a listener method. Submit and ImageSubmit
have a listener parameter that may be bound to a listener method, just like a
DirectLink or Form component. This allows the previous example to be rewritten more simply:
listener="ognl:listeners.moveUp"
label="Move Up"/>
listener="ognl:listeners.moveDown"
label="Move Down"/>
120
CHAPTER 3
In this example, if the user clicks the first button, the moveUp() listener method
will be invoked. If the user clicks the second button, the moveDown() listener
method will be invoked. If the form is submitted otherwise, neither listener
method is invoked. In either case, the form’s listener is invoked last. The Submit
component’s listener method is invoked just as the component renders itself
(during the Form’s rewind phase). This means that not all properties that will be
set during the form submission will have been updated yet.
It is even possible to combine the two approaches, in which case the selected
parameter is updated before the listener method is invoked.
TIP
The listener method must take care not to upset the rewind. Remember
that the rewind must go through the exact same steps as the initial render, or the rewind will terminate with a StaleLinkException. A Submit
component can trigger this by changing properties of the page haphazardly. The most common example is a Submit component used in a
form, enclosed within a Foreach. The Submit components should not
change the list used as the source parameter of the Foreach; doing so
will affect the remaining iterations through the Foreach loop. Instead,
the listener methods should note what operation should take place and
store this information in temporary page properties. From the Form’s
listener, which is invoked only after the rewind completes, the desired
operation can take place.
Repeatedly, we’ve referenced the idea that Form components can enclose
Foreach components. We’ll next explore an example of a complex form that
combines several types of basic form control components within a Foreach.
3.5 Creating a to-do list
Now that we’ve introduced the basic Tapestry form components, let’s do something just a little more ambitious than the login form. Let’s create a simple onepage to-do list, as shown in figure 3.5.
This kind of application should be familiar to anyone who has used a personal digital assistant (PDA). There is a list of entries, which may be prioritized by
the user. New entries can be created at any time, and existing entries may be
marked as completed. When desired, completed entries may be removed from
the list.
Figure 3.5 shows a single form, in which you could conceivably mark an item
as done, change the title of another, and change the order of the items all as a
single request.
121
Figure 3.5
Users can create
new entries, update
and delete entries,
and reorder the
entries in the list.
We’ll be making use of Checkbox, TextField, and Submit components in the
application and (as promised) show how to use loops inside a Form.
3.5.1 Defining the data object
The first step is to identify how the data edited by the page will be stored. Each
item in the list has three properties: a title, a completed flag, and its position in
the overall list. We’ll have the item store the completed flag and the title directly,
and make the list responsible for item position. Listing 3.5 contains the source
for this class.
Listing 3.5 ToDoItem.java: the data object used in the ToDo list page
package examples.todo1;
import java.io.Serializable;
public class ToDoItem implements Serializable
{
private boolean _completed;
private String _title;
public ToDoItem()
{
}
public ToDoItem(String title)
{
_title = title;
}
122
CHAPTER 3
public boolean isCompleted()
{
return _completed;
}
public String getTitle()
{
return _title;
}
public void setCompleted(boolean completed)
{
_completed = completed;
}
public void setTitle(String title)
{
_title = title;
}
}
This class is simply a property holder for two properties: completed and title. A
more sophisticated application would track more data, such as priority level or
due date, but this is sufficient for this example. Because instances of this class
will be stored persistently (as persistent page properties ultimately stored in the
HttpSession), the class implements the Serializable interface. This is necessary
to allow the application server to support clustering and failover (a subject discussed in greater detail in chapter 7).
With the domain object for this application created, the next step is to define
the user interface for editing a list of these items.
3.5.2 Creating the ToDo HTML template
The template for the ToDo page consists of a Form. Within the Form is a Foreach
that loops over the ToDo items, and additional components for editing the properties of each item in the list. Listing 3.6 shows this template.
Listing 3.6 ToDo.html: the HTML template for the ToDo page
<html jwcid="@Shell" title="ToDo List">
<body>
<form jwcid="@Form" listener="ognl:listeners.formSubmit">
<table>
<tr>
<th>Done</th>
<th>Title</th>
<th colspan="2">Reorder</th>
</tr>
<tr jwcid="@Foreach" element="tr"
source="ognl:toDoList"
value="ognl:item">
<td><input type="checkbox"
jwcid="completed@Checkbox"
selected="ognl:item.completed"/></td>
<td><input type="text"
jwcid="title@TextField"
size="50"
value="ognl:item.title"/></td>
<td><input type="submit" jwcid="up@Submit"
selected="ognl:moveUpItem"
tag="ognl:item"
label="Move Up"/></td>
<td><input type="submit" jwcid="down@Submit"
selected="ognl:moveDownItem"
tag="ognl:item"
label="Move Down"/></td>
</tr>
</table>
123
b
c
d
<input type="submit" value="Update"/>
<input type="submit" jwcid="add@Submit"
listener="ognl:listeners.addTodoItem" label="Add Item"/>
<input type="submit" jwcid="deleteCompleted@Submit"
listener="ognl:listeners.deleteCompleted"
label="Delete Completed"/>
e
</form>
<hr/>
<p><a href="#" jwcid="@PageLink" page="Home">Return to
Home page</a>.</p>
</body>
</html>
b
This loops through the list of items supplied by the toDoList property, assigning
each in turn to the item property. In addition, specifying a value for the element
parameter causes the Foreach to render a <tr> element around its body for each
loop, allowing the component to blend into the structure of the HTML <table>
seamlessly and invisibly.
124
CHAPTER 3
c
If the user clicks the Move Up button, the current item is assigned to the moveUpItem property; the actual change in ordering occurs inside the Form’s listener
method, formSubmit().
d
Alternately, the user may click the Move Down button, and the item is assigned
to the moveDownItem property.
These two buttons are outside the loop. Each invokes a listener method
(addTodoItem() or deleteCompleted()) when clicked.
e
Here we’re making use of another feature of the Foreach component: the element parameter. When a value is bound to the element parameter, then on each
loop, the Foreach component renders the specified element around its body. In
this example, the Foreach component will not only render the form element
components it encloses; it will also render a <tr> element around each set of
those form elements.
Notice that the Foreach component is positioned within the template using a
<tr> element; in this way, the component integrates seamlessly into the overall
template. Alternately, we could use a <span> tag as we have in the past with the
Foreach component:
<span jwcid="@Foreach" source="ognl:toDoList"
value="ognl:item">
<tr>
. . .
</tr>
</span>
This alternate usage will produce virtually identical HTML output (with minor
variations in whitespace) but is more intrusive; for example, it puts a <span> tag
directly inside a <table> element, which is not valid HTML.
The page’s HTML template shows another variation on the jwcid attribute:
named implicit components. Putting a component ID before the @ symbol
assigns a specific ID to the component. In this example, the Checkbox component has the jwcid completed@Checkbox. This is still an implicit component
(nothing more about the component goes in the page’s specification); it just has
the ID completed rather than a framework-assigned ID.2 In a bit, we’ll provoke
some errors related to these components, and the error messages will be more
intelligible because we used named, rather than anonymous, components.
2
As you’ve seen in previous examples, anonymous components are assigned IDs by the framework,
based on the type of component. In this case, the component ID would have been $Checkbox.
125
Much like the Hangman examples in the previous chapter, the page becomes
the mediator between the components. The Foreach component gets a list of
ToDoItems from the page and updates the item property for each ToDoItem within
that list. Within the Foreach, the other components (the Checkbox, the TextField, and the two Submit components) are editing the current item within the
loop, or other properties of the page based on the current item.
Outside the Foreach loop are additional Submit components that trigger the
other two behaviors: adding a new ToDoItem, and deleting any items that are
marked completed. The next step is to create the page specification, including
additional properties for item and toDoList.
3.5.3 Specifying properties in the page specification
All the components in the page are fully defined in the HTML template, but a
page specification (provided in listing 3.7) is still needed for two reasons: to specify
the Java class for the page and to specify any properties Tapestry should create.
Listing 3.7 ToDo.page: specification for the ToDo page
<page-specification class="examples.todo1.ToDo">
<property-specification name="toDoList"
type="java.util.List"
persistent="yes"/>
<property-specification name="item"
type="examples.todo1.ToDoItem"/>
<property-specification name="moveUpItem"
<property-specification name="moveDownItem"
The first property, toDoList, is the list of ToDoItems on the page. This property is
persistent; it will be stored between requests and restored in each subsequent
request. Persistent page properties are covered in more detail in chapter 7.
The item property is the property set by the Foreach component on each pass
through its loop. The other two properties, moveUpItem and moveDownItem, are set
by the corresponding Submit components. These two components don’t change
126
CHAPTER 3
the toDoList property, since that would upset the Form rewind. Instead, the two
properties are set by the components during the rewind and are later accessed
from inside the Form’s listener method.
3.5.4 Initializing the toDoList property
Because the property type for toDoList is java.util.List, a collection type, it is
awkward to specify its initial value in the page specification. We want to set the
toDoList property to a reasonable value just as it is needed. Once again, we
implement the PageRenderListener interface and provide a pageBeginRender()
implementation to initialize the toDoList property the first time it is needed
(just before the page renders for the first time). The ToDo page class includes
the following code to perform this initialization:
public abstract List getToDoList();
public abstract void setToDoList(List toDoList);
{
List list = getToDoList();
if (list == null)
{
list = new ArrayList();
list.add(createNewItem("Finish reading Tapestry Book"));
list.add(createNewItem("Download latest version of Tapestry"));
setToDoList(list);
}
}
protected ToDoItem createNewItem(String title)
{
return new ToDoItem(title);
}
When the ToDo page is first rendered, the toDoList property will be null. In this
case, a new ArrayList is created, and two initial items are added to it. The property is then updated by invoking the setToDoList() method, and (because the
page specification indicates that toDoList is a persistent property) behind the
scenes, Tapestry stores the list into the HttpSession for use in later requests. This
value will be used in the current request when the page renders but will also be
available later, when the form within the page is submitted.
The Checkbox and TextField components do all the work of editing each
ToDoItem’s completed and title properties. Managing the order of items inside
the list requires a bit of support by the page, which we look at in the next section.
127
3.5.5 Handling reordering
Moving a to-do item up or down within the list requires just a bit of care. It is
important that the Submit components not change the toDoList property while
the form is still rewinding; this could lead to properties of the wrong objects getting updated. That’s the first principle of Tapestry forms: The rewind phase
must match the original render, object for object.
Instead, each of the two Submit components, when triggered, must note
which item is to be moved up or down so that the reordering can occur later
when it is safe to change the list. We’ve already declared two properties for this
purpose in the page specification. We’ve also ensured that the properties are
updated by making use of the selected and tag parameters of the two Submit
components in the HTML template. The item property is set by the Foreach
component before it renders the portion of the template containing the TextField, Checkbox, and Submit components. Each Submit component determines,
as it rewinds, whether it was the cause of the form submission (it will see a nonnull query parameter matching its element ID). If so, it updates its selected
parameter from its tag parameter and then invokes its listener.3
The Form component’s listener method is invoked after the rewind completes, when it is safe to change the toDoList property. This is where the items
are actually moved. The formSubmit() listener method, from the ToDo page
class, is the right place to make all these changes:
public void formSubmit(IRequestCycle cycle)
{
List list = getToDoList();
int count = list.size();
ToDoItem moveUpItem = getMoveUpItem();
ToDoItem moveDownItem = getMoveDownItem();
{
ToDoItem item = (ToDoItem) list.get(i);
if (item == moveUpItem)
{
if (i > 0)
Collections.swap(list, i, i - 1);
3
All of these parameters are optional. A Submit (or ImageSubmit) simply doesn’t set its selected
parameter if the parameter is not bound, and it doesn’t invoke a listener unless one is provided.
128
CHAPTER 3
break;
}
if (item == moveDownItem)
{
if (i + 1 < count)
Collections.swap(list, i, i + 1);
break;
}
}
setToDoList(list);
}
Notice that, after modifying the list, the code invokes the method setToDoList()
again. Tapestry relies on the developer to inform it when properties are changed
by invoking the accessor method. This usually isn’t an issue, because the typical
types for persistent properties (numbers, booleans, and strings) are immutable.
For mutable types, like List, Tapestry will not know to refresh the copy of the
property value stored in the HttpSession unless the setter method is invoked.
This subject is covered in greater detail in chapter 7.
3.5.6 Deleting completed items
The Submit component labeled Delete Completed is used to delete from the list
all to-do items that have been marked completed. Unlike the reordering Submit
components, this component’s listener is invoked outside the Foreach component, so it is safe for its listener to modify the toDoList property directly. The listener for this component is the deleteCompleted() listener method of the ToDo
page class:
public void deleteCompleted(IRequestCycle cycle)
{
ListIterator i = getToDoList().listIterator();
while (i.hasNext())
{
ToDoItem item = (ToDoItem) i.next();
if (item.isCompleted())
i.remove();
}
}
We do take one liberty here: We should invoke setToDoList() at the end of the
method. This can be skipped only because the Form’s listener, the method
129
formSubmit() listed earlier, is invoked after this listener method and always
invokes setToDoList().
3.5.7 Triggering stale links
There is an inherent problem in basing the dynamic portion of a form (such as
the list of items in the ToDo page) on persistent properties: It is possible for the
web browser and the server to get out of synchronization with each other. When
the form is submitted, Tapestry will detect that the incoming form does not
match the state stored on the server, a situation known as a stale link. When Tapestry detects a stale link, it throws a StaleLinkException. Having the application
fail in this way is not acceptable in a production application, and we discuss how
to avoid stale links in chapter 4. In the meantime, it is useful to understand the
underlying causes of stale links. We can trigger a stale link for the ToDo application easily by following these steps:
1
Start the application fresh.
2
Go to the ToDo page.
3
Click the Add Item button.
4
Press the browser’s back button (so that just two items are visible).
5
Click the Update button.
You should see the StaleLink page, as shown in figure 3.6. This page includes a
message to help you figure out what went wrong: Rewind of form ToDo/$Form expected
allocated id #9 to be ‘add’, but was ‘completed$1’ (requested by component ToDo/completed).
What’s happened is that the form submitted with two items, but the rewind had
three items to iterate over (including the new item added before you clicked the
browser’s back button). At the point that the completed component (a Checkbox)
rendered for the third time, a mismatch occurred because when the page was originally rendered, there were only two items in that loop. The first form element after
the loop is the add component (a Submit). That’s where the mismatch occurs; the
data recorded in the submitted form indicates that the next allocated element ID
should be add, but the form’s rewind includes a third pass through the loop and
wants to allocate a third ID for the completed component.4
4
The Form adds suffixes as necessary to form unique element IDs, starting with the element’s ID as a
base. So on the first pass through the loop, the completed component got the element ID completed,
on the second pass it got the element ID completed$0, and on the third pass, completed$1.
130
CHAPTER 3
Figure 3.6 The StaleLink page is triggered by submitting a request where the client no longer
matches server-side state.
Always give IDs to your form control components, especially if you are
doing any looping or conditionals inside the Form; otherwise, the message on the StaleLink page will be much less intelligible, since the element IDs will be assigned by the framework.
TIP
Doing a view source on the page before submitting the form helps show what
happened:
<form method="post" name="Form0" action="/examples/app">
<input type="hidden" name="service" value="direct/1/ToDo/$Form"/>
<input type="hidden" name="sp" value="S0"/>
<input type="hidden" name="Form0"
value="completed,title,up,down,completed$0,title$0,
up$0,down$0,add,deleteCompleted"/>
b
➥
<table>
<tr>
<th>Done</th>
<th>Title</th>
</tr>
<tr>
<td><input type="checkbox" name="completed"/></td>
<td><input type="text" name="title"
c
value="Finish reading Tapestry Book" size="50"/></td>
<td><input type="submit" name="up" value="Move Up"/></td>
<td><input type="submit" name="down" value="Move Down"/></td>
</tr>
<tr>
<td><input type="checkbox" name="completed$0"/></td>
<td><input type="text" name="title$0"
value="Download latest version of Tapestry" size="50"/>
</td>
<td><input type="submit" name="up$0" value="Move Up"/></td>
<td><input type="submit" name="down$0" value="Move Down"/></td>
</tr>
</table>
<input type="submit" name="add" value="Add Item"/>
<input type="submit" name="deleteCompleted"
value="Delete Completed"/>
</form>
b
c
d
e
131
c
d
e
The Form component writes this hidden field, recording all the IDs allocated for
all the enclosed components. It uses this information when the user submits the
form to check that the same sequence of IDs is allocated during the rewind.
This portion of the output HTML is written during the first pass through the
Foreach component’s loop. The element IDs for the components inside the loop
are completed, title, up, and down.
This portion of the output HTML is written during the second pass through the
loop. The element IDs for the components are completed$0, title$0, up$0, and
down$0. The $0 extension is to make the element IDs unique within the Form.
The server-side state mandates that a third pass through the loop should take
place here (because there are three to-do items), but the form was rendered when
there were only two items in the list.
In the output HTML, we can see how the Form component uses hidden fields to
store the list of form element IDs. This list is used during the rewind phase to
ensure that the render and the rewind do, in fact, match.
Because there are three items stored in the toDoList property of the page,
there should be a third <tr> within the table, containing another Checkbox, a TextField, and pair of Submit buttons. This is discovered by the Form during the
rewind phase, at which point it throws the StaleLinkException (which results in
the Stale Link page being displayed).
We’ve shown you how to trigger a stale link under controlled circumstances.
You can be assured that if you aren’t prepared for it, your users will find ways to
trigger stale links under uncontrolled circumstances—typically by clicking their
132
CHAPTER 3
web browser’s back button. Your end users should never see an exception page or
a Stale Link page, so in chapter 4 we discuss additional components and techniques that detect and avoid stale links.
3.6 Summary
Tapestry relieves you of all the drudgery of both rendering forms and form elements and processing form submissions. A non-Tapestry servlet application will
contain reams of tedious code for reading query parameters and converting
string values. When you use Tapestry, this code simply disappears, leaving just
higher-level code that operates directly on the properties of the domain objects.
As in the examples with simple links and images, Tapestry pages with forms still
look and feel like ordinary HTML pages; the Tapestry extensions fit invisibly into
place. Once again we’ve seen where Tapestry’s simplicity and consistency goals
come into play: It’s far easier to use the Form component and form control components to read and update properties of the page directly than it ever would be
to fuss with query parameters.
Even when adding the complication of editing a list of items within a form,
Tapestry performs well. Despite a few constraints placed on the developer by the
framework, the effort and code required to process such complex forms is still
minimal. That is far from all the framework has to offer in terms of forms support, however. In the following chapter, we pull even further ahead and describe
advanced form components that do much more than simply mimic ordinary
HTML elements.
Advanced form
components
This chapter covers
■
Complex components vs. simple components
■
Creating drop-down lists
■
Storing hidden data in forms
■
Uploading files
133
134
CHAPTER 4
In chapter 3, we covered the basic set of Tapestry components—components that
map directly to HTML form control elements. These basic components provide a
baseline for form-related functionality, but Tapestry is capable of much more. In
this chapter, we’ll introduce more sophisticated components that do more: more
HTML, more client-side JavaScript, more server-side processing. Much of this
chapter will focus on reworking the ToDo application from chapter 3, adding
features and, ultimately, addressing the browser back button issue. In addition,
we’ll introduce components for uploading files and for selecting dates using a
pop-up calendar. Chapter 5 will discuss the input-validation subsystem provided
by Tapestry.
4.1 Introducing the advanced form components
The previous chapter covered a number of Tapestry components used for basic
HTML form support. Each of the basic components maps directly to a particular
HTML form control element. This chapter looks at components that break that
mold by adding significant capabilities in terms of client-side or server-side processing (or both). These components, shown in table 4.1, are more task oriented
than the element-oriented components of the previous chapter.
Table 4.1
Component
Description
PropertySelection
Creates a drop-down list of values
Hidden
Records a property for a hidden form field and restores it on submission
ListEdit
Iterates over a list of items like Foreach, but works better within a form
Upload
Allows files to be uploaded from the client workstation
DatePicker
Uses a JavaScript pop-up window to let the user enter a date
The first of these components is the widely used PropertySelection component,
which provides a streamlined approach to creating single-selection drop-down lists.
4.2 Creating drop-down lists with PropertySelection
A common feature in many web forms is a drop-down list that lets you make a
selection. Whether you’re asking users to choose their state of residence, a type
of credit card, or some other information that has a limited set of options, it is
often easiest for you, and for your users, to let them select from a list rather than
Creating drop-down lists with PropertySelection
135
prompt them to type in a value. A drop-down list makes economical use of the
available screen real estate. As you’ll recall, chapter 3 describes the Select and
Option components, which are one approach to creating drop-down lists.
Using the Select and Option components for this purpose, however, is awkward. The Option component reads and sets a boolean property, and it requires
you to implement supporting code in the page to translate between property values and a set of boolean flags. You saw this in the examples in chapter 3, where
we used a synthetic boolean property to control the adding of items to or removing of items from a set. Select and Option are focused on handling multiple
selection, not the more common case where only a single selection from a dropdown list is desired. Fortunately, we aren’t restricted to using only Select and
Option components when creating drop-down lists.
The PropertySelection component exists specifically for the purpose of
allowing a user to select a single option from a drop-down list. PropertySelection
will create a <select> element and a set of <option> tags when rendering, and
will perform an update of a property when the form is submitted. You must
provide the PropertySelection component with information about the options
that are allowed. To do this, you bind the PropertySelection’s model parameter
to an object implementing the IPropertySelectionModel interface. The PropertySelection component uses the model when rendering and again when the
form is submitted.
The model provides the following information to PropertySelection:
■
The total number of options
■
The property value for each option
■
The label for each option (which is displayed to the user)
■
The encoded value for each option, used as the value attribute of the
<option> element
■
The property value corresponding to an encoded value
This is an example of the Model-View-Controller pattern in action, even though
it involves only a single component within a page rather than an entire page.
The PropertySelection component does not have any special knowledge about
the options it presents to the user; it will work with any model object provided to
it. The IPropertySelectionModel interface is the contract between the model and
the component.
The PropertySelection component is quite flexible. A simple implementation
of the model supports the selection of a string value from a list (the property
136
CHAPTER 4
edited by the PropertySelection component will be a string). At the other
extreme, an application may use a PropertySelection component to edit a relationship between two database objects. For example, an order-management form
within an e-commerce application’s account management page may allow you to
select an active order from a drop-down list so that you can review the order’s
progress. Here, the option label will be a string identifying the Order object, such
as a confirmation ID number. The property value will be the Order object itself,
and the encoded option value will be the Order object’s primary key.
In keeping with Tapestry’s goal of consistency, you can use a PropertySelection
component for both simple and complex models. To demonstrate how to use
PropertySelection and how to create a model for it, let’s look at an improved
version of the ToDo application.
4.2.1 Adding priority levels to the ToDo application
In this version of the to-do list, each item stores a priority value, along with its
title and completed flag. The priority can be Low, Medium, or High. Figure 4.1
shows the new version of the page.
Figure 4.1
The improved ToDo page, with a PropertySelection for choosing the priority for each item.
137
As before, our first step is to identify how this new data will be stored. For flexibility, we’ll store the priority as an integer, using 100 for Low, 200 for Medium,
and 300 for High. At a later date, we can then easily add new values between
these initial values (say, adding “Important” as 250). Because we’ll use the other
existing properties of the ToDoItem class, let’s create a subclass that adds a new
property for storing the priority value. This new subclass is shown in listing 4.1.
As you can see, the ToDoItem2 class includes public constants used to define the
three priority values.
Listing 4.1 ToDoItem2.java: data object for the ToDo2 page
import examples.todo1.ToDoItem;
public class ToDoItem2 extends ToDoItem
{
public static final int LOW_PRIORITY = 100;
public static final int MEDIUM_PRIORITY = 200;
public static final int HIGH_PRIORITY = 300;
private int _priority = MEDIUM_PRIORITY;
public ToDoItem2()
{
}
public ToDoItem2(String title)
{
super(title);
}
public int getPriority()
{
return _priority;
}
public void setPriority(int priority)
{
_priority = priority;
}
}
138
CHAPTER 4
4.2.2 Updating the HTML template
The next step is to create a new version of the ToDo page that includes a PropertySelection component for editing the priority properties of the to-do items. Let’s
start with a copy of the template from the earlier version of the ToDo list page
and add new HTML and components to support the new property. The end
result is the HTML template in listing 4.2.
Listing 4.2 ToDo2.html: HTML template for the ToDo2 page
<html jwcid="@Shell" title="ToDo List (version 2)">
<body>
<table>
<tr>
<th>Done</th>
<th>Priority</th>
<th>Title</th>
</tr>
<tr jwcid="@Foreach"
element="tr" source="ognl:toDoList"
value="ognl:item">
<td><input type="checkbox" jwcid="completed@Checkbox"
selected="ognl:item.completed"/></td>
<td><select jwcid="priority@PropertySelection"
The PropertySelection
value="ognl:item.priority"
component
model="ognl:priorityModel"/></td>
<td><input type="text" jwcid="title@TextField" size="50"
value="ognl:item.title"/></td>
<td><input type="submit" jwcid="up@Submit"
selected="ognl:moveUpItem"
tag="ognl:item"
label="Move Up"/></td>
<td><input type="submit" jwcid="down@Submit"
selected="ognl:moveDownItem"
tag="ognl:item"
label="Move Down"/></td>
</tr>
</table>
<input type="submit" jwcid="add@Submit"
listener="ognl:listeners.addTodoItem"
label="Add Item"/>
<input type="submit" jwcid="deleteCompleted@Submit"
listener="ognl:listeners.deleteCompleted"
label="Delete Completed"/>
139
</form>
<hr/>
<p><a href="#" jwcid="@PageLink" page="Home">Return to
Home page</a>.</p>
</body>
</html>
The only significant change to the original HTML template is the addition of the
PropertySelection component. Its value parameter is bound to the property we
want to edit, just as with a TextField. Here, the PropertySelection component will
edit the current item’s priority property. Unlike with a TextField, the property
may be of any type, not just a string.
4.2.3 Implementing the page class
For the most part, the ToDo2 class is the same as for the original page. There are
two differences:
■
ToDo2 provides a read-only property, priorityModel, needed by the Property-
Selection component.
■
When adding new items, we instantiate the ToDoItem2 class instead of ToDoItem.
Because they have so much in common, ToDo2 subclasses the original class, ToDo,
but adds one new method and overrides another. First, the priorityModel property (referenced in the HTML template by the PropertySelection component) is
provided via the getPriorityModel() method. The model is needed by the PropertySelection component to create the drop-down list of priorities:
public IPropertySelectionModel getPriorityModel()
{
return new PriorityModel();
}
New item instances are created in only one place: the createNewItem() method.
Overriding this method in the ToDo2 subclass allows for the creation of the
ToDoItem2 class:
{
return new ToDoItem2(title);
}
140
CHAPTER 4
All the remaining behavior from the first page applies to this page as well, and is
simply inherited from the original page class.
4.2.4 Implementing the model
The PropertySelection component’s model is a class that implements the IPropertySelectionModel interface, which consists of five methods. These methods define
all the options that will be presented in the drop-down list.
When rendering, PropertySelection creates a <select> element, and then
uses the model to obtain the number of options. The PropertySelection component interacts with the model during the render, as shown in figure 4.2. For each
option provided by the model, PropertySelection obtains the option value (which
may be an object type), the string value (which is the value encoded into the
form as the <option> element’s value attribute), and the label that will be displayed to the user. This information is used to write each <option> element
Figure 4.2 The property selection model provides the number of options as well as the information
needed to render each option.
141
within the <select>. The option value is used to determine whether the <option>
should be selected by default.
Listing 4.3 contains the PropertySelection model for selecting priority levels.
It defines a class, PriorityModel, that implements the IPropertySelectionModel
interface—the interface a PropertySelection model class must implement.
Listing 4.3 PropertyModel.java: PropertySelection model implementation
import org.apache.tapestry.form.IPropertySelectionModel;
public class PriorityModel implements IPropertySelectionModel
{
public int getOptionCount()
{
return 3;
}
public Object getOption(int index)
{
switch (index)
{
case 0 :
return new Integer(ToDoItem2.HIGH_PRIORITY);
case 1 :
return new Integer(ToDoItem2.MEDIUM_PRIORITY);
default :
return new Integer(ToDoItem2.LOW_PRIORITY);
}
}
public String getLabel(int index)
{
switch (index)
{
case 0 :
return "High";
case 1 :
return "Medium";
default :
return "Low";
}
}
b
Converts
index to
property
value
142
CHAPTER 4
public String getValue(int index)
{
return Integer.toString(index);
}
c
Converts index
to option value
public Object translateValue(String value)
{
int index = Integer.parseInt(value);
return getOption(index);
d
Converts option
value to property
value
}
}
b
c
d
The getOptionCount() method returns 3, because there are always exactly three
options (High, Medium, and Low).
The getValue() method returns the option value recorded into the form (as the
value attribute of the <option> element). For this model, the option value is simply the index into the list of possible options—a common, efficient approach
when the list of options is fixed at compile time.
The translateValue() method is the counterpart to getValue(). It translates the
option value back to an index number and returns the option for that index.
The first method in the IPropertySelectionModel interface, getOptionCount(),
provides the number of items in the list. For this model, there are always exactly
three options—High, Medium, and Low—so getOptionCount() always returns 3.
The PropertySelection component will iterate through the possible indexes (0
through 2) and invoke the next three methods for each index.
The PropertySelection component invokes the getOption() method only
when it is rendering (the method isn’t invoked when the containing form is submitted). This method allows the PropertySelection component to determine
which option, if any, should be selected initially. The first option whose value (as
returned by getOption()) matches the value from the component’s value parameter is selected. (The <option> element will include the selected attribute.) Keep
in mind that although this example edits an integer property, PropertySelection
has no knowledge of the type of property it is editing; that information is entirely
contained within the model. The value parameter could just as easily be a double, a string, a date, or a custom application object—even an object retrieved
from a database. The return value for the getOption() method is Object, and the
three priority level constants are wrapped as Integer objects and returned.
The second method, getLabel(), provides the label for each option. As with
getOption(), the parameter is the index within the loop. The returned label is
143
presented directly to the user. In an internationalized application, the label value
returned must be localized to match the page’s locale.
As the PropertySelection component renders the <option> elements, it must
have a string to use as the value attribute. The PropertySelection component
allows the model to specify this in the getValue() method. A common approach,
which we use here, is to encode the index as a string. The model should be able
to convert this string back to an option value when the form is submitted.
In this particular model, we could have encoded the value for one of the
three constants here. That is, we could return 100 , 200 , or 300 rather than
the index.
The final model method, translateValue(), is used only when the form is
submitted. The PropertySelection component will use this method to translate
an encoded value directly to an option so that the option can be used to update
the property bound to the PropertySelection’s value parameter. The translateValue() method is the counterpart to the getValue() method. Figure 4.3 shows
how this method is used by the PropertySelection component.
For this model implementation, it is convenient to simply convert the index
value back to an integer and use the getOption() method to obtain the value to
be assigned.
Figure 4.3 The PropertySelection rewind sequence. The value is extracted from the query parameter and
the model is used to translate it to a value, which is then assigned back to the domain object property.
144
CHAPTER 4
By combining the PropertySelection component with a custom model class,
we have easily created a working drop-down list in our application with little
more effort than would be required when using a TextField. With some minor
changes to our Java code, we can even get rid of the custom model class and use
an existing model implementation provided by the framework. We just have to
replace our integer constants with enums.
4.2.5 Using enums instead of integers
Using a set of integers to represent the priority levels is somewhat of an oddity in
Java coding style. In C or C++, you would almost certainly use an enum instead.
Enums are preferable to integers because they limit the possible values to a finite
set, which is exactly what we want here. The compiler can do much better type
checking of method parameters or other expressions for enums than it can for
simple integers. Java currently does not have an enum construct. Sun is expected
to add support for enums in Java Development Kit (JDK) 1.5; but until that version of Java is widely available, you can simulate C enums by using Java classes.
In this take on the ToDo list application, we’ll switch from representing the
priority as integers and instead use an enumerated type. We’ll also make use of a
framework-provided model class, tailored for use with enumerated types. The
final application looks identical to the previous version, ToDo2. It’s simply a different way of storing data on the server side.
One of the advantages of building and using open-source software is the
availability of other high-quality frameworks and libraries. Tapestry makes use of
a number of such frameworks, including the OGNL library, as well as other
frameworks available from the Jakarta project that hosts Tapestry. The Jakarta
commons-lang library contains a number of handy utilities and classes, one of
which is an Enum base class used for creating enumerated types. Creating a new
enumerated type is as simple as creating a subclass of Enum and defining some
public constants for the possible values. Listing 4.4 shows the implementation of
the Priority class.
Listing 4.4 Priority.java: enum type used with the ToDo3 class
import org.apache.commons.lang.enum.Enum;
public class Priority extends Enum
{
public static final Priority HIGH = new Priority("HIGH");
145
public static final Priority MEDIUM = new Priority("MEDIUM");
public static final Priority LOW = new Priority("LOW");
public static final Priority[] ALL_VALUES =
{ HIGH, MEDIUM, LOW };
private Priority(String name)
{
super(name);
}
}
The constructor for Priority is private; this means that the class can’t be
extended. There’s no way to create additional instances of Priority beyond the
three values defined here: Priority.HIGH, Priority.MEDIUM, and Priority.LOW
(except by changing the Priority class to define new values). If we ever write
code that needs to differentiate ToDoItem3 instances based on priority, we can
make direct comparisons to these three constant values; that is, we can use
the Java equality operator, ==, rather than invoke the equals() method. The
Enum base class ensures the identity of instances, even when a Priority is serialized and deserialized,1 such as when the value is directly or indirectly stored in
the HttpSession.
From there, we again extend the ToDoItem class to define a priority property
of type Priority. The new class, ToDoItem3, appears in listing 4.5.
Listing 4.5 ToDoItem3.java: data item used with the ToDo3 page
public class ToDoItem3 extends ToDoItem
{
private Priority _priority = Priority.MEDIUM;
public ToDoItem3()
{
}
1
Normally, when an object is serialized and then deserialized, the result is a copy of the original object—
a new and separate instance. This makes sense, because (for a while) the object was just a stream of
bytes. The Enum base class includes the necessary hooks to find the correct singleton after deserializing, which ensures that, within a JVM, there is exactly one instance of any particular Enum value (such
as Priority.HIGH).
146
CHAPTER 4
{
super(title);
}
public Priority getPriority()
{
return _priority;
}
public void setPriority(Priority priority)
{
_priority = priority;
}
}
The HTML template for the ToDo3 page is identical the ToDo2 HTML template
(except for the page title). There’s nothing to change; PropertySelection is still
editing the priority property of the item, and the page still provides an instance
of IPropertySelectionModel to handle all the translations as a priorityModel
property. On the implementation side, the type of the priority property in the
item class has changed from int to Priority, and the model used by the PropertySelection component has changed as well. But none of that is relevant to the
interface or even to the PropertySelection component—it’s all encapsulated
inside the model. This is a good example of the power of the Model-ViewController pattern: The View (the HTML template) is unchanged even when the
Model (the ToDoItem3 class) is altered considerably; the Controller (the Java page
class) has modest changes to fulfill its role of bridging the two.
The ToDo3 page specification differs only in the page class we want to instantiate. The createNewItem() method is overridden in the ToDo3 class to instantiate
an instance of ToDoItem3. The only real change is related to creating a model for
the PropertySelection component—this appears in the ToDo3 page class:
private IPropertySelectionModel _priorityModel;
public IPropertySelectionModel getPriorityModel()
{
if (_priorityModel == null)
_priorityModel = buildPriorityModel();
b
Builds and caches
the model
return _priorityModel;
}
private IPropertySelectionModel buildPriorityModel()
{
ResourceBundle bundle =
ResourceBundle.getBundle("examples.todo3.PriorityStrings",
getLocale());
Constructs the model
c
return new EnumPropertySelectionModel(Priority.ALL_VALUES,
bundle);
Uses the correct
}
b
c
d
147
d
locale for the page
The model, once constructed, is immutable. It is constructed the first time it is
needed and then saved for later requests.
The model is constructed using the properly localized ResourceBundle and the
list of possible values (in the order they should appear in the drop-down list).
This will be a file, PriorityStrings.properties, in the same package folder as the
Java class.
The page returns, in its locale property, the locale that should be used for any
visible text.
The core of this approach is the Tapestry framework class EnumPropertySelectionModel. This is an implementation of IPropertySelectionModel geared around
Enums. The first parameter for the constructor is an array of the Enum values, in
the order in which they should appear in the drop-down list created by the PropertySelection component.
The second parameter is a ResourceBundle containing the localized labels we
want to use.2 The keys in the ResourceBundle are the names of the Enum instances,
in this case HIGH, MEDIUM, and LOW. The PriorityStrings.properties file contains the
user-presentable labels:
HIGH=High
MEDIUM=Medium
LOW=Low
Tapestry does quite a bit in terms of managing the end user’s locale; we cover
this topic in detail in chapter 7. For now, suffice to say that the page knows the
correct locale for any localizable resources such as the PriorityStrings.properties
file, and this is accessible via the getLocale() method inherited from the
AbstractPage superclass.
At runtime, the PropertySelection component will use the EnumPropertySelectionModel to access the localized strings in the ResourceBundle. The model
2
ResourceBundle is a utility class provided by the JDK. A ResourceBundle is a container of string keys
and values, typically read from a set of files. ResourceBundles are used primarily to support localiza-
tion; they allow literal strings to be moved out of Java code and into separate files, which may be edited
and translated individually.
148
CHAPTER 4
also will be responsible for creating option values and translating option values
to one of the three options: Priority.HIGH, Priority.MEDIUM, or Priority.LOW.
The end user will not see any difference in the behavior of the ToDo2 and
ToDo3 pages, but internally, the implementation of the ToDo3 page is cleaner
and simpler.
The PropertySelection component vastly simplifies the effort needed to create a drop-down list. This single component can be used to create lists for anything from a simple Yes or No, to a list of strings, all the way up to a list of objects
read from a database. You simply provide the model listing the available options
and tell the component how to store this information in the HTML form, as well
as how to convert it back to application data.
In the next two sections, we describe Tapestry form components that are
invisible to the end user. They allow information to be stored in the HTML form
as hidden input fields and permit that information to be retrieved when the user
submits the form.
4.3 Recording data in the form with Hidden
One of the more vexing problems in web application development is how to deal
with the browser back button. In a traditional desktop application, the user
interface displayed to the user is always in perfect synchronization with the running application, but in a web application this isn’t always so. Because the user
can hit the browser back button to return to a previous page (really, a previous
rendering of a page stored in a cache within the web browser) and continue to
click links and submit forms, it is all too easy for the user interface to be out of
synch with the state of the application running on the server.
For example, consider a typical e-commerce application with a product catalog. A user may start a search for, say, digital cameras. The user eventually
reaches a product details page for a particular camera, perhaps a Minolta. From
there, the user clicks on a Related Items link and gets the product details for a
similar Nikon camera.
Unsatisfied with the Nikon, the user hits the browser back button and
returns to the page displaying the Minolta camera. This Minolta page will
come out of the browser’s page cache, and no interaction with the application
server takes place. To the user, these are two different and distinct pages, one
showing the Minolta and one showing the Nikon, exactly as if the user was
thumbing through a print catalog. To the application, these are different configurations of the same page, rendered at different times. Unlike a desktop
Recording data in the form with Hidden
149
application, a web application can only guess at what is on users’ screens when
they submit forms.
The user then decides to purchase the Minolta and clicks the Add To Shopping Cart button to add the camera to the cart. Which camera gets added to the
cart: the Minolta or the Nikon? It depends on how the application is coded.
The application may store the identity of the product displayed on the product details page on the server. Tapestry includes a facility for this (persistent page
properties), but even a non-Tapestry application could do this by storing an
attribute into the HttpSession. If the application operates this way, the user will
be surprised by a Nikon camera in the shopping cart, not the Minolta the user
was expecting. Because the server has no way of knowing that the user hit the
browser’s back button, it can only assume that the user is looking at the most
recently rendered page (the one showing the Nikon details), even if the user is
instead looking at the Minolta camera in a page cached within the browser.
We need a way to encode in the form the identity of the camera so that when
the form is submitted, the correct camera is added to the shopping cart. In traditional web application development, this is accomplished by including a hidden
field in the form. You create a hidden field by using the HTML element <input
type="hidden">. A hidden field works like a text field, except that no user interface element is created in the web browser. The hidden field is, simply, hidden,
and the value for the field is submitted along with the other fields of the form.
Tapestry includes a component, Hidden, which builds on the HTML hidden
form control. The Hidden component works much like a TextField component;
when it renders, it reads a domain object property bound to its value parameter.
Unlike with TextField, the property does not have to be a string; this property
type is encoded along with its value, just as you’ve seen before with DirectLink
component parameters. Figure 4.4 shows the sequence when a Hidden component renders.
The Form component is responsible for writing the hidden field (as an <input
type="hidden"> element). In most cases, the domain object is the page object
itself; as with all component parameters, it doesn’t matter to the Hidden component where the value comes from. When the form is submitted, the value is
extracted from the request, converted back into an object type, and used to
update the domain object property, as shown in figure 4.5. A listener method
may optionally be specified; if so, the listener is invoked after the property is
updated. Remember that listener methods are really IActionListener instances.
Invoking the listener method gives the page a chance to perform any additional
synchronization operations related to the data stored in the hidden field.
150
CHAPTER 4
Figure 4.4 When the Hidden component renders, it extracts a property and
records it in the form as a hidden field.
Figure 4.5 After reading and converting the query parameter, the Hidden component updates its
value parameter and, finally, invokes its listener.
151
In the e-commerce example, the Hidden component would record just the identity
of the Product entity for the camera. This might be a SKU or a database primary
key. The listener method, invoked after the id property is restored, can then go
to the database to read the Product entity and continue with the process of adding the camera to the shopping cart.
We provide a more complete example of using the Hidden component in
chapter 10.
The Hidden component is useful for recording a single value in a form. However, there are cases where an entire list of values should be encoded. In the next
section, we show you how to properly loop over such a list of values inside a form.
4.4 Looping within a form using ListEdit
Dealing with lists is a common issue related to synchronization between the client and the server. Forms may contain a Foreach component and iterate over a
list of values, allowing properties of each element in the list to be edited. This is
common in e-commerce applications, where a “shopping cart” page allows quantities of all the items in the shopping cart to be edited in one place. The same
concept applies in many other types of applications as well.
As you saw in chapter 3, it is easy for users to trip up the Form component by
using their browser’s back button, triggering a StaleLinkException. For simple
forms, without loops or conditionals, this is never an issue. However, for complex
forms, which include loops and conditionals, it is possible for a Form submission
to get out of sequence with the previous Form render. The simplest way for a synchronization fault to occur is when the form includes a loop that iterates over
data that may change between the time the page is rendered and the time the
form on the page is submitted. This often occurs when a loop works off data
from a database, and rows are added to, or removed from, the database after the
page is rendered but before the form is submitted (for example, if a second user is
updating the database). A similar scenario is one in which the user backtracks to
a page that was rendered when the application was in a different state.
So, just as the Hidden component can store a single value in a form, the
ListEdit component can store a list of values in a form—and iterate over them
just like a Foreach component. The interface for ListEdit resembles that of the
Foreach component; it also has source and value parameters. When a ListEdit is
rendering, it behaves exactly like the Foreach component, with one difference: It
records a series of hidden fields in the form, one for each element in its source
list. As with the Hidden component, it encodes each value, maintaining its type
for later, when the form is submitted. Figure 4.6 shows how this works.
152
CHAPTER 4
Figure 4.6
When rendering, the
ListEdit component acts
much like a Foreach.
The listener is invoked
after the value is
updated on each pass
through the loop.
A major departure from the Foreach component is the existence of the ListEdit’s
listener parameter. Whereas a Foreach component updates its value parameter
and renders its body, the ListEdit component updates its value parameter and
then invokes a listener method before rendering its body. The intent is that the
successive values are some form of object ID, and the listener method is supposed to read the full object for that ID and make it available as a page property.
When the enclosing Form is rewinding, ListEdit works very differently from
Foreach. During the rewinding, a ListEdit component reads the hidden fields it
previously recorded (during the render); it doesn’t use its source parameter at
all when rewinding. The ListEdit component still updates the property bound
to its value parameter, just as a Foreach component does, before rendering its
body. The operation sequence for rewinding is just about the same as for rendering (except for the interaction with the Form component to record hidden
field values).
153
While the ListEdit component provides the functionality for iterating over
lists, it is commonly combined with a helper class to fill in the details, as you’ll
see in the next section.
4.4.1 Using the ListEditMap
The ListEdit component provides the raw structure for dealing with HTML form
synchronization issues, but it requires a bit of effort to use. There’s a lot of work
related to supplying ListEdit with a list or collection of IDs, converting those IDs
to objects, and so forth. That can turn into a bit of code, so, naturally, Tapestry
offers some assistance to help you avoid unnecessary coding.
The framework includes a companion class, ListEditMap, which provides a
server-side mapping from the object IDs that are stored in the form to the objects
for those IDs, which are available on the server. A page must create a ListEditMap
instance and load it with the object IDs and matching object values before rendering the ListEdit component, and then must reconstruct the ListEditMap
instance when the form is submitted. The ListEdit component’s source and
value parameters are connected to ListEditMap’s keys and key properties, as
shown in table 4.2.
Table 4.2
ListEditMap properties
ListEdit parameter
ListEditMap property
Usage
source
keys
The collection of keys to iterate over (render only)
value
key
A value set by ListEdit to the current key within the
list of keys (render and rewind)
value
The value corresponding to the current key (render
and rewind)
With the ListEdit component properly configured, we can extend the ListEdit
sequence diagram to show exactly how ListEditMap is involved, as shown in figure 4.7.
The listener method can get the current value from the ListEditMap instance.
This listener method should always check to see if the value is null; this can
occur when two users’ updates overlap. If the first user deletes some objects, the
second user’s form submit will contain the IDs of deleted objects, which will not
be present in the ListEditMap instance (because they have been deleted from the
underlying database).
To see how ListEdit and ListEditMap work together, let’s update our ToDo
application again, replacing Foreach with a ListEdit component.
154
CHAPTER 4
Figure 4.7 The ListEdit component is connected to the ListEditMap instance and
keeps it informed about the current object ID (by setting the key property). The
listener method can easily obtain the current object for that key from ListEditMap
via the value property.
4.4.2 Using ListEdit in the ToDo application
In chapter 3, we demonstrated how easy it is to force the ToDo application
(which is based on a Foreach component) to fail (with a StaleLinkException) by
using the browser’s back button. Now we’ll create a fourth version of the application that doesn’t have that limitation.
Creating a ToDo item with an ID
The ListEditMap class is built around the idea that items have some form of
unique key that can be used to identify them. In a production application, these
155
keys are likely to be some form of database primary key. The ToDoItems used in
these example applications are not stored in a database, but we can still create a
kind of unique key for the items. The first step is to extend the ToDoItem class yet
again. Listing 4.6 shows the implementation of ToDoItem4, which extends from
ToDoItem3 and adds a unique immutable identifier, as an int, to each item.
Listing 4.6 ToDoItem4.java: data object for the ToDo4 page
import examples.todo3.ToDoItem3;
public class ToDoItem4 extends ToDoItem3
{
private static int _nextuid = 0;
private int _uid = _nextuid++;
b Source of unique IDs
unique ID
c Assigns
when object created
public ToDoItem4()
{
super();
}
{
super(title);
}
public int getUid()
{
return _uid;
}
}
b
c
A simple static variable is used to allocate unique IDs. In a real application,
unique IDs would be provided by a database. This approach is acceptable only
for this simple demonstration.
The unique ID is assigned as each instance is constructed.
Now that we have to-do items that can be identified using a unique ID, we can
start updating the application to make use of the ListEdit component, starting
with the page template.
156
CHAPTER 4
Updating the page template
The template for the ToDo4 page is for the most part identical to the ToDo3 page,
except for the one line concerned with the item loop:
<tr jwcid="listEdit" element="tr">
Because the ListEdit component has even more parameters than the Foreach
component used in the previous examples, it makes sense to configure those
parameters in the ToDo4 page specification:
<component id="listEdit" type="ListEdit">
<binding name="source" expression="listEditMap.keys"/>
<binding name="value" expression="listEditMap.key"/>
<binding name="listener" expression="listeners.synchronizeItem"/>
</component>
Whereas the Foreach component in the previous to-do list examples read the
toDoList property and directly updated the item property, here the ListEdit
component reads and updates properties of the ListEditMap class. The item
property still gets updated, but indirectly, by the synchronizeItem() listener
method. The page is responsible for initializing the listEditMap property before
the page renders. The ListEditMap is initialized with the values from the toDoList property (you’ll see the details of this shortly) and acts as a buffer between
the ListEdit component and the properties of the page.
In addition, the ToDo4 page may need to display an error message if a form
submission cannot be processed; the HTML template includes a familiar pair of
components for displaying the error message (if the message is non-null):
<span jwcid="@Conditional" condition="ognl:errorMessage">
<span style="{ font: bold; color: red }">
<span jwcid="@Insert" value="ognl:errorMessage">Error Message
</span>
<br/>
</span>
</span>
As in previous examples, this HTML template snippet checks to see if there is a
non-null error message and, if so, displays it at the top of the page, before the
form renders.
Specifying the page properties
The ToDo4 page uses the same set of specified properties as the ToDo3 page,
with two additions: a listEditMap property (connected to the ListEdit component) and an errorMessage property (used to store an error message display
157
when the form submission cannot be processed due to synchronization problems). All of these properties are defined in the ToDo4 page specification:
<property-specification name="toDoList"
type="java.util.List" persistent="yes"/>
<property-specification name="item"
<property-specification name="moveUpItem"
<property-specification name="moveDownItem"
<property-specification name="listEditMap"
type="org.apache.tapestry.form.ListEditMap"/>
<property-specification name="errorMessage"
type="java.lang.String"/>
Now that we’ve specified the necessary properties, we need to make sure they are
initialized and used properly.
Initializing the ListEditMap
Listing 4.7 contains the code for the ToDo4 page class. It extends the ToDo3 class,
adding support for initialization and using the ListEditMap instance.
Listing 4.7 ToDo4.java: page class for the ToDo4 page
import java.util.List;
import
import
import
import
org.apache.tapestry.PageRedirectException;
org.apache.tapestry.form.ListEditMap;
import examples.todo3.ToDo3;
public abstract class ToDo4 extends ToDo3
{
public abstract void
setErrorMessage(String message);
public abstract void
setListEditMap(ListEditMap listEditMap);
public abstract ListEditMap getListEditMap();
public abstract void setItem(ToDoItem item);
Accessors for
specified
properties
{
return new ToDoItem4(title);
Creates new items as
}
instances of ToDoItem4
158
CHAPTER 4
b
{
super.pageBeginRender(event);
Invoked when page
renders or when
form rewinds
ListEditMap map = new ListEditMap();
List items = getToDoList();
int count = items.size();
{
ToDoItem4 item = (ToDoItem4) items.get(i);
int uid = item.getUid();
map.add(new Integer(uid), item);
}
setListEditMap(map);
}
c
Sets up
toDoList
property
Adds item to the
ListEditMap
Stores the map
for later
public void synchronizeItem(IRequestCycle cycle)
{
ListEditMap map = getListEditMap();
ToDoItem item = (ToDoItem) map.getValue();
if (item == null)
{
setErrorMessage(
"Your form submission is out of date. Please retry.");
throw new PageRedirectException(this);
}
➥
d
Is invoked
by ListEdit
setItem(item);
}
}
b
c
d
The pageBeginRender() method will be invoked when the page renders or when
a form within the page rewinds. The super implementation must be invoked,
because that ensures that the toDoList property is initialized.
Each item is added to the ListEditMap, which remembers the order in which key/
value pairs are added.
The ListEdit component invokes this listener method after setting ListEditMap’s
key property. The map’s getValue() method returns the corresponding object,
as previously stored inside pageBeginRender(). If the current key is not stored in
the map, then null is returned—this can happen when a user submits a form
after clicking the back button.
159
The toDoList property is the list of items, in the correct order (the order as
manipulated by the user). ListEditMap remembers the order that items are
added to it; its getKeys() method returns the keys in that order. It is safe to cast
each item in the list to ToDoItem4, since the createNewItem() method has been
overridden to instantiate this class.
Synchronizing the item
The second responsibility of the class is to synchronize the item property—the
ListEdit component stores a list of to do item IDs in the form, and the page class
must use these to set the item property to the correct instance of ToDoItem4.
Nearly all the work for this is done by ListEditMap with a short listener method,
synchronizeItem() (shown in listing 4.7), which gets the current item from the
ListEditMap and assigns it to the page’s item property. In addition, the synchronizeItem() method includes a check for a null item.
As we’ve described earlier, the ListEdit component will set the key property of
the ListEditMap before invoking the listener method. Invoking getValue() on
the ListEditMap returns the corresponding value object, as previously recorded
into the map; this is an instance of ToDoItem4. The check for null covers race
conditions between two users,3 or one user using the browser’s back button.
Throwing a PageRedirectException aborts the form’s rewind phase entirely and
forces the page to render as is.
Chapter 10 describes some further functionality provided by the ListEditMap
class—specifically, the ability to track values within the map that should be
deleted. The ToDo application doesn’t use this feature of ListEditMap, because it
has its own approach for deleting completed items in the list.
The previous two sections described components, Hidden and ListEdit, that
are critical to the infrastructure of an application but largely invisible to end
users. Let’s now return to components that users will see and interact with.
3
A race condition is a category of software bug that is concerned with multiple users affecting the same
data simultaneously. In a race condition, two (or more) users attempt the same operation at roughly
the same time, and the first user to complete the operation affects the outcome of the other users. A
common example of a race condition in a web application is when one user deletes an object stored
in a database and another user attempts to update the same object. Whether the second user’s update
succeeds or fails depends on which user wins the race.
160
CHAPTER 4
4.5 Handling file uploads
A common feature of many web applications is support for file uploads, the ability to transfer a file from the user’s computer to the application server as part of
a form submission. HTML includes another form control for this purpose, <input
type="file">. When a web browser submits a request with a file upload, the normal encoding used to express the data (the Multipurpose Internet Mail Extensions [MIME] type application/x-www-form-urlencoded) is not used. Instead, a
different encoding, multipart/form-data, is used for the upload.
This alternate encoding allows any number of file uploads, interspersed with
normal query parameter values (from the other form controls), to be sent from
the client to the server in one large binary stream. This difference in encoding
should not be any more of an issue to a servlet application developer than the
difference between the HTTP GET and PUT requests. Alas, the Servlet API does
not include the ability to parse and interpret multipart/form-data content,
which makes handling uploaded files an uphill battle.
Tapestry makes use of the Jakarta commons-fileupload library to seamlessly
support forms that contain a mix of ordinary form elements and file uploads.
When a form is submitted containing an uploaded file, Tapestry extracts the file
content from the request and stores it in memory or in a temporary file (if the
file content is large enough).
To the application, an uploaded file is represented by an instance of the interface IUploadFile. From this, an application can retrieve the following information:
■
The name of the file (on the client)
■
The complete path of the file (on the client)
■
The MIME content type (as reported by the client)
■
The content of the file, as a java.io.InputStream
The content of the file is deleted at the end of the request; so if the uploaded file
is to be used later, it must be stored persistently either to the server’s file system
or in a database.
Tapestry’s Upload component makes creating file upload fields as easy as creating ordinary text fields. Figure 4.8 shows a page that uses an Upload component.
It is just as easy to add an Upload component to a Tapestry Form as any
other kind of component. The HTML for the Upload component from figure 4.8
is simply
<input type="file" jwcid="@Upload" file="ognl:file" size="100"/>
161
Figure 4.8 A page containing an Upload component. Clicking the Browse button will open a file
selection dialog box.
Including an Upload component inside a Form automatically changes the encoding type of the enclosing Form to multipart/form-data. This is another example
of the dynamic nature of Tapestry; you are not responsible for this background
detail (selecting the form’s encoding type)—the components automatically do
the right thing.
Once the form is submitted and the file is uploaded, the file instance can be
passed to another page in the same way as a simple value, such as a string or
number. Inside the Upload page class, the listener method does just that:
public abstract IUploadFile getFile();
{
IUploadFile file = getFile();
if (file == null)
return;
UploadResults next =
(UploadResults) cycle.getPage("UploadResults");
next.setFile(file);
cycle.activate(next);
}
The second page, UploadResults, displays the data available about the
uploaded file, followed by a dump of the contents of the file (in hexadecimal
162
CHAPTER 4
Figure 4.9 After the file is uploaded, the application displays the information available about the
uploaded file.
and ASCII ). Figure 4.9 shows an example of what the page looks like after
uploading a small file.
Getting the output for the first three fields (name, path, and content type)
is straightforward:
<table border="0">
<tr>
<th>Name:</th>
<td><span jwcid="@Insert" value="ognl:file.fileName">
File Name</span></td>
</tr>
<tr>
<th>Path:</th>
<td><span jwcid="@Insert" value="ognl:file.filePath">
File Path</span></td>
</tr>
<tr>
<th>Content Type:</th>
<td><span jwcid="@Insert" value="ognl:file.contentType">
text/html</span></td>
</tr>
163
Getting the output for the binary content is a bit more involved. Tapestry doesn’t
have a built-in component for this kind of output, but it does include a utility
class, BinaryDumpOutputStream. BinaryDumpOutputStream is a filter that takes as
input a stream of bytes and produces as output the kind of text shown in
figure 4.9. The trick is to get this object, which is not a component, integrated
into the rendering of the page. Fortunately, as you’ve seen before, components
aren’t the only things that can render output in Tapestry. The UploadResults
page class in listing 4.8 includes an inner class that can render the contents of
an IUploadFile.
Listing 4.8 UploadResults.java: page class for the UploadResults page
package examples.upload;
import
import
import
import
java.io.IOException;
java.io.InputStream;
java.io.OutputStream;
java.io.StringWriter;
import
import
import
import
import
import
import
org.apache.tapestry.ApplicationRuntimeException;
org.apache.tapestry.IMarkupWriter;
org.apache.tapestry.IRender;
org.apache.tapestry.request.IUploadFile;
org.apache.tapestry.util.io.BinaryDumpOutputStream;
public abstract class UploadResults extends BasePage
{
public abstract void setFile(IUploadFile file);
public abstract IUploadFile getFile();
private static class ContentRenderer
implements IRender
{
private IUploadFile _file;
b
Renders binary
output
ContentRenderer(IUploadFile file)
{
_file = file;
}
public void render(IMarkupWriter writer,
IRequestCycle cycle)
{
try
{
StringWriter buffer = new StringWriter();
c
Is defined by the
IRender interface
164
CHAPTER 4
BinaryDumpOutputStream out =
new BinaryDumpOutputStream(buffer);
out.setBytesPerLine(32);
out.setShowAscii(true);
InputStream in = _file.getStream();
d
Gets content of
uploaded file
copy(in, out);
in.close();
out.close();
writer.print(buffer.getBuffer().toString());
}
catch (IOException ex)
{
throw new ApplicationRuntimeException(
"Unable to generate binary output.", ex);
}
}
private void copy(InputStream in, OutputStream out)
throws IOException
{
byte[] buffer = new byte[1000];
while (true)
{
int length = in.read(buffer);
if (length < 0)
return;
out.write(buffer, 0, length);
}
}
}
public IRender getContentRenderer()
{
return new ContentRenderer(getFile());
}
}
b
An inner class is defined to render the binary content of the file as a hexadecimal
dump. The IRender interface is the common interface for any kind of object,
component or not, that can be part of the page-rendering process.
c
d
165
The render() method is the sole method defined by the IRender interface.
The uploaded file is an instance of IUploadFile. The getStream() method provides a binary input stream of the uploaded content in the file.
To get this output, we make use of a Delegator component. Delegator delegates
its rendering to another object, an object that implements the IRender interface.
The rendering object is specified using the delegate parameter. The HTML template includes the Delegator component, which delegates to the contentRenderer property of the page:
<pre>
<span jwcid="@Delegator" delegate="ognl:contentRenderer"/>
</pre>
All that’s left is to provide the contentRenderer property on the page:
public abstract IFile getFile();
public IRender getContentRenderer()
{
return new ContentRenderer(getFile());
}
When the page renders, the Delegator component invokes the render() method
on the ContentRenderer instance created by this method; ContentRenderer is
responsible for reading the content of the uploaded file and outputting that content, formatted by the BinaryDumpOutputStream instance, into the response.
The FileUpload component shows how a Tapestry component can perform
considerable server-side processing. The next component, DatePicker, shows the
kind of client-side logic that can be provided by a component.
4.6 Creating pop-up date selections using DatePicker
Last but not least on our tour of advanced Tapestry form components is the
DatePicker component. DatePicker allows users to enter dates using a sophisticated pop-up calendar window. The component generates client-side JavaScript
that allows a pop-up window to be displayed when the user clicks an “activate”
button to the right of the text field. Alternately, the user may type directly into
the text field. Figure 4.10 shows a DatePicker component in action.
DatePicker allows the user to navigate easily to other months and years using
the drop-down controls; it also allows the user to move forward or backward a
month at a time using the two buttons in the upper-left and upper-right corners
of the window. The current date may be selected by clicking the large button
166
CHAPTER 4
Figure 4.10 The DatePicker component uses client-side JavaScript to create a
pop-up window for selecting dates.
along the bottom of the window, and clicking the Clear button closes the window
with no date selected.
Using a DatePicker component is as easy as using any other Tapestry
component (with one caveat, which we discuss in a moment), despite the fact
that DatePicker has a considerable client-side presence. Listing 4.9 shows all
that’s necessary.
Listing 4.9 DatePicker.html: HTML template for DatePicker page
<html jwcid="@Shell" title="Date Picker">
<body jwcid="@Body">
Enter a date: <input jwcid="@DatePicker"
value="ognl:date"/>
DatePicker
component
<br/>
<input type="submit" value="Submit"/>
</form>
167
<p><a href="#" jwcid="@PageLink" page="Home">
Return to Home page</a>.</p>
</body>
</html>
Most of this template should look familiar. The DatePicker component has a
value parameter bound to the date property of the page, and the DatePicker
component, being a form control component, is enclosed by a Form component.
When the page is rendered, DatePicker outputs a text field and the trigger button (to the right of the text field) that displays the calendar pop-up window.
The one caveat is the use of a Body component for the <body> element of the
template. The Body component organizes all the client-side JavaScript on the
page into a single <script> block, just inside the <body> tag. Tapestry components, such as DatePicker, that use client-side JavaScript expect to be enclosed by
a Body component. The Body component is also utilized to generate unique
names for variables and JavaScript functions, which allows any number of
DatePicker components to be used within a page. We’ll discuss creating components with a client-side JavaScript aspect in more detail in chapter 8.
TIP
If you fail to wrap a DatePicker component within a Body component,
Tapestry will throw an exception: DatePicker components must be contained
by a Body component. Using a Body component doesn’t cost any measurable amount; you should just get in the habit of always using one.
WARNING
Because of a bug in Microsoft Internet Explorer, drop-down lists will always render on top of the DatePicker pop-up; this is because the z-order
of drop-down lists is fixed at the highest possible value. When creating
the layout for a page using the DatePicker component, you must take
this into consideration and ensure that the DatePicker component’s
pop-up window will display in an area free of drop-down lists.
Again, the simplicity of Tapestry is apparent here, since the DatePicker component
is as easy to use, if not easier, than a TextField component. It’s the same pattern:
pick a component, place it in the HTML template, and configure its parameters.
The fact that this component is internally complex is of no concern to the page
using the component or to you as the developer when constructing the page.
168
CHAPTER 4
4.7 Summary
Tapestry’s more advanced form-related components are all built around usability: both for you, the developer, and your end users. Tapestry components not
only encapsulate presentation issues, but also handle the conversions between
domain properties and the string values used within HTML. The framework’s
component-based approach to form handling makes it possible to build forms
with a much higher level of complexity without getting bogged down in the
mundane details of string conversions and field naming. Once again, Tapestry
lets you get more work done, better and more quickly—which leaves you more
time to concentrate on the specifics of your application.
The next step is to integrate these basic form-handling concepts into an overall approach to validating user input. As you’ll learn in the next chapter, Tapestry has a complete subsystem for just this purpose.
This chapter covers
■
Knowing the requirements of a usable,
validating interface
■
Using validators and validation delegates
■
Using FieldLabel and ValidField components
■
Adapting output for your application’s look
and feel
■
Performing form-level validations
169
170
CHAPTER 5
Processing forms is more than creating drop-down lists and text fields; the forms
are just part of an overall cycle of user input and server validation. Users will
make mistakes when entering data into forms. They’ll leave a required field
blank, enter letters into fields clearly marked for numeric input, ignore any kind
of range requirement, and type whatever they want. It isn’t enough just to reject
invalid input from the user as an exception case; users can be expected to find
creative ways to enter invalid input on even the simplest forms. Your responsibility is to handle invalid input consistently and gracefully—that is, to create a
usable application. Fail to take your users’ needs and expectations into account by
creating a slipshod, unfriendly application and you run the risk of frustrating
them—and driving them away from your site. Users have some basic expectations when it comes to input validation:
■
Fields with invalid data will be marked as such; whether this is with color,
icons, or some other approach, it should be possible to identify fields with
problems at a glance. An unfriendly application will just display an error
message and expect the user to deduce the invalid field.
■
Invalid input should be maintained so that it can be corrected. For example, if a user types only 15 digits of a credit card number, you should give
the user a chance to enter just the 16th digit. An unfriendly application
forces the user to retype the entire number from scratch.
■
All errors should be visible at once, to allow the user to correct all the
errors without additional server requests. An unfriendly application can
recognize only one error at a time, forcing the user to submit the form
multiple times.
■
Client-side validations are preferable to server-side validations.
Tapestry includes an entire subsystem for validating user input, centered on its
ValidField component. The subsystem allows you to easily build highly usable
forms that provide useful feedback and client-side validation. We’ll begin by
demonstrating how to use validation for a simple user registration form, and
later, we’ll show you how to mix and match the validation subsystem with the
DatePicker component from chapter 4.
5.1 Validating user input
The ValidField component is a variation of the TextField component described
in chapter 3. For the most part, a ValidField component is used in exactly the
same way as a TextField component—with some additional parameters. ValidField
171
components are capable of editing not just string properties, but dates and
numbers as well, and can be adapted to any data type with a reasonable text
representation. The validations that can be used with ValidField components
can apply both client-side and server-side checks on input. These checks
ensure that the user provides values for required fields, or force user input to
be within a specified range. In many cases, the validations are tied in with
conversions, such as converting input from a string to a date or to some type
of number.
An example of ValidField components in use appears in figure 5.1, which
shows a page that accepts a user’s name and address—the kind of page you’ll see
in many online applications. Initially, all the fields are blank, and the cursor is
Figure 5.1 The Register page in its initial state. Each required field is marked with an asterisk, and
the cursor is placed into the first required field.
172
CHAPTER 5
automatically placed into the first required field (using a snippet of client-side
JavaScript, automatically generated by the ValidField component).
If you enter values into some fields but not others and submit the form (this
form does not have client-side validation enabled), you’ll see a different screen
(figure 5.2). The page is redisplayed so that you may make corrections. Details
about the first field error appear at the top; each field that is in error is so
marked in several ways: the label for the field, and the field itself, both change
color, and an error marker is added to each field.1
Figure 5.2 The same form partially filled out and submitted. The first error on the page is displayed
prominently. All fields with errors are highlighted in three ways: the field label text is red, the field
itself gets a red background, and an icon is displayed to the right of the field.
1
As you’ll see, you can easily customize the look and feel for validation to fit seamlessly into your application.
173
So, how does this all work? It’s more than what a single component in isolation can accomplish; a component can normally affect only a small portion of
the overall page, but figure 5.2 reflects changes to the output HTML scattered
throughout the entire rendered page. Making this work requires one additional
component (FieldLabel) and two additional objects (a validator that parses and
validates the user input, and the validation delegate, which tracks fields and
errors), with some subtle interactions between all four.
The next few sections give a quick overview of the various parts of the validation subsystem.
5.1.1 Using FieldLabels in conjunction with ValidFields
A FieldLabel is a companion component to a ValidField. Each FieldLabel is connected to ValidField via the FieldLabel’s field parameter. The FieldLabel can
adjust its visual look to reflect the field. If the field is in error (because of invalid
user input), the FieldLabel can display itself differently. Figure 5.2 shows this; the
labels on several fields have been marked red to indicate that the fields are in error.
In addition, the FieldLabel obtains the user-presentable name of the field to
display from the field itself (ValidField has a displayName parameter used for this
purpose). This ensures that the field label matches the name of the field used in
any error messages, even when the name of the field is localized into the user’s
language, or when the field name is determined dynamically.
FieldLabels and ValidFields are largely responsible for presenting the interface
to the user; the heart of input validation is provided in specialized validator objects.
5.1.2 Using validators
Validators are objects that are responsible for translating user input from strings
into object types (such as numbers or dates) and performing other validation
checks. Validators are not components themselves; they are objects that implement the IValidator interface. Validators are used by ValidField components,
which delegate all the conversions and checks to the validator. A validator object
has four responsibilities:
■
Converting an object value (from a domain object property) to a string
that can be used when rendering the page. The converted value is used
as the value attribute of the <input> element rendered by the ValidField component.
■
Converting a string value submitted with the form back into an object
value so that it can be used to update a domain object property.
174
CHAPTER 5
■
Performing additional validations on the input. These validations are controlled by properties of the validator instance, which you must configure.
For example, you may set the maximum property of a NumberValidator to
enforce a maximum allowable value.
■
Writing any client-side JavaScript needed to perform client-side validations.
Each ValidField component has a validator parameter, which is bound to the
validator object for that field. Validator objects are meant to be shared; they
don’t store any information about a particular ValidField. Many ValidFields can
share the same validator instance.
All validators include a required property of type boolean. When the
required property is set to true, a validator will not allow an input field to be
blank or to consist only of whitespace.
Tapestry provides a number of implementations of the IValidator interface
that can be used and configured off the shelf. StringValidator is used for editing string properties of pages or domain objects and can apply an additional validation: a minimum number of characters that you want the field to accept.
NumberValidator is used for editing numeric properties of all types (int, long,
BigDecimal, and so on). NumberValidator can enforce a minimum or a maximum
value, or both. DateValidator is used for editing Date properties. As with NumberValidator, you can set a minimum or a maximum value. Later in this section,
you’ll see how to create a custom validator, one that enforces the format of a
postal zip code (which can be seen in action in figure 5.3).
The final piece of the validation puzzle is the validation delegate: an object that
is used to track which fields are in error within a form, and what error(s) are
associated with each field.
5.1.3 Using validation delegates
The validation delegate has two distinct functions. First, it tracks the error state
of each ValidField enclosed by a form. When the form is submitted, each ValidField passes the string provided by the user to the validator object for that
field. The validator object may convert the string to an object type, such as
Long or Double, or leave it as a string. It will also apply any validations, such as
checking that the converted value fits into a specified range. If the converted
value fails a validation, the validator reports the error back to the ValidField.
The conversions and validations a validator can perform are flexible, and, as
you’ll see, it is easy to create new validators to handle new types of conversions
and validations.
175
Figure 5.3 A custom validator attached to the Zip field knows the correct format for zip codes and
can generate a custom error message.
Validators report these errors by throwing a ValidationException. The ValidField catches the exception and uses the validation delegate to record the ValidField’s element ID, the exception message, and the invalid input provided by the
user. The recorded exception message is the indicator that the field is in error,
and will be used when the page containing the form is rendered.
Validation delegates have a second, discrete function. The delegate is responsible for decorating the fields and labels that are in error. The validation delegate
has opportunities to render additional HTML before and after the FieldLabel
renders, and before and after the ValidField renders, and can even write additional attributes into the <input> element rendered by the ValidField itself. The
exact sequence for the FieldLabel component is shown in figure 5.4.
176
CHAPTER 5
Figure 5.4 The field label gets the user-presentable name for the field from the ValidField
component. It allows the delegate to render before and after it renders the field name, so
that the delegate can decorate the label if the corresponding field is in error.
This additional rendering is how the labels for invalid fields, shown in figure 5.2,
manage to be displayed in red. The validation delegate has a chance to wrap the
FieldLabel’s output in a <span> element, and the span references a Cascading
Style Sheet (CSS) class that results in the label being displayed in red. The methods writeLabelPrefix()and writeLabelSuffix() allow the validation delegate to
decorate the label by writing extra HTML around it. Likewise, both the validator
and the validation delegate are integrated into the render of the ValidField, as
The validator’s renderValidatorContribution() method is primarily used by
the validator to write client-side JavaScript (which will perform client-side validations). The validation delegate methods (primarily, the writeAttributes()
177
Figure 5.5 The delegate and the validator are both hooked into the rendering of the ValidField.
The validator may use its hook to create client-side validation JavaScript. The delegate uses its
hooks to decorate fields that are in error.
method) are used to decorate the <input> element (rendered by the ValidField) if
the field is in error.
The primary purpose of creating a custom subclass of ValidationDelegate is
to override these rendering methods so that an application-specific look and feel
for decorating fields and labels can be created. In the examples in this chapter,
we’ve done just that to mark the fields and labels in red, and to add an error icon
after invalid fields.
178
CHAPTER 5
The validator objects and the validation delegate object are needed by the
Form and ValidField components. They are connected to the components using
the components’ parameters (the delegate parameter of the Form component,
and the validator parameter of the ValidField component). These objects don’t
appear out of thin air; they must be instantiated and configured. This can be
done in Java code (within your page class) or by using another feature of the
page specification: helper beans.
5.1.4 Using helper beans
As we’ve been discussing, handling validation is about more than pages and
components. It involves the participation of at least two other objects: the validator attached to each ValidField, and the validation delegate that tracks field
input errors throughout a form.
Both of these objects are JavaBeans—but where and when do they get instantiated?
One possibility is to make instantiation the responsibility of the page. The
page class could include fields and accessor methods, such as
private IValidator _required;
public IValidator getRequired()
{
if (_required == null)
{
_required = new StringValidator();
_required.setRequired(true);
}
return _required;
}
However, that kind of mechanical coding goes against the grain of Tapestry. Getting these helper beans instantiated and configured falls into the general “plumbing” category—the category of things that should be shifted into the framework.
As you’ve seen before, when Tapestry provides an alternative to coding, it involves
the page (or component) specification. In this case, we want to be able to control
which Java class is instantiated and how it is configured. This is accomplished in
the page specification using the <bean> element:
<bean name="required"
class="org.apache.tapestry.valid.StringValidator"
lifecycle="page">
<set-property name="required" expression="true"/>
</bean>
Building the Register page
179
The <bean> element shown here is equivalent to the Java code snippet. The
<bean> element specifies three things:
■
The Java class you want to instantiate
■
The configuration of any properties of the helper bean
■
The lifecycle of the bean
A helper bean may be accessed through the beans property of the page or component, using its name. For example, the OGNL expression beans.required
would access the bean specified here. Helper beans are created only as needed,
the first time the bean is referenced.
By default, the lifecycle of a bean is request, meaning that the bean will be
released to the garbage collector at the end of the current request cycle. This is
often appropriate, especially if the bean contains any state particular to the current user—such information should last only as long as the current request so
that it won’t be visible in a subsequent request from a different user. By using the
page lifecycle, the bean is kept for as long as the page instance exists; this is
appropriate for beans such as a validator that have no internal state. As you’ll see
in the following examples, a common use for helper beans is to define validators
and validation delegates.
So far, you’ve seen an overview of how the various aspects of the validation
subsystem (the components, the validators, and the validation delegate) operate.
Next let’s return to the Register page (shown in figure 5.1) and start seeing how
these bits and pieces fit together to form a working, validated form.
5.2 Building the Register page
The ultimate aim of the Register page is to collect a user’s name and address and
store it in an Address (shown in listing 5.1).
Listing 5.1 Address.java: data object used with the Register page
package examples.register;
public class Address implements Serializable
{
private String _firstName;
private String _lastName;
private String _address1;
private String _address2;
180
CHAPTER 5
private String _city;
private String _state;
private String _zip;
public String getAddress1()
{
return _address1;
}
public String getAddress2()
{
return _address2;
}
public String getCity()
{
return _city;
}
public String getFirstName()
{
return _firstName;
}
public String getLastName()
{
return _lastName;
}
public String getState()
{
return _state;
}
public String getZip()
{
return _zip;
}
public void setAddress1(String address1)
{
_address1 = address1;
}
public void setAddress2(String address2)
{
_address2 = address2;
}
public void setCity(String city)
{
181
_city = city;
}
public void setFirstName(String firstName)
{
_firstName = firstName;
}
public void setLastName(String lastName)
{
_lastName = lastName;
}
public void setState(String state)
{
_state = state;
}
public void setZip(String zip)
{
_zip = zip;
}
}
Armed with this definition of what goes into an address, we can create a user
interface to let users enter an address.
5.2.1 Creating the Register HTML template
The HTML template for the Register page is shown in listing 5.2. This template
introduces a number of new concepts, so we’ll take it apart one small piece at a time:
■
Using the delegate parameter of the Form component
■
Using the page’s components property to reference other components
within the page’s template
■
Using the FieldLabel component in conjunction with the ValidField component
Listing 5.2 Register.html: HTML template for the Register page
<html jwcid="@Shell" title="Registration"
stylesheet="ognl:assets.stylesheet">
<head jwcid="$remove$">
<link rel="stylesheet" type="text/css"
href="css/style.css"/>
</head>
b
Elements
marked for
removal
182
CHAPTER 5
<span class="title">Registration</span>
<p>Please enter your mailing address for our records.
Fields marked with a
<span class="required-marker">*</span>
are required.
</p>
<span jwcid="@Conditional"
condition="ognl:beans.delegate.hasErrors">
<table class="error">
<tr valign="top">
<td>
<img height="52" alt="[Error]" src="images/form-error.png"
width="52">
</td>
<td>
<span jwcid="@Delegator"
delegate="ognl:beans.delegate.firstError">
Error Message
</span>
</td>
</tr>
</table>
</span>
<form jwcid="@Form"
listener="ognl:listeners.formSubmit"
delegate="ognl:beans.delegate">
c
Validation delegate
for the Form
<table class="form">
Labels connected
<tr>
to ValidFields
<th><span jwcid="@FieldLabel"
field="ognl:components.inputFirstName">First Name
</span></th>
<td><input type="text" jwcid="inputFirstName" size="50"/></td>
</tr>
d
ValidField as a
declared component
<tr>
field="ognl:components.inputLastName">Last Name
</span></th>
<td><input type="text" jwcid="inputLastName" size="50"/></td>
</tr>
<tr>
field="ognl:components.inputAddress1">Address
e
</span></th>
<td><input type="text" jwcid="inputAddress1" size="50"/></td>
</tr>
<tr>
<td></td>
<td><input type="text" jwcid="@TextField"
value="ognl:address.address2" size="50"/></td>
</tr>
<tr>
field="ognl:components.inputCity">City
</span></th>
<td><input type="text" jwcid="inputCity" size="50"/></td>
</tr>
<tr>
field="ognl:components.inputState">State
</span></th>
<td><input type="text" jwcid="inputState" size="2"/></td>
</tr>
<tr>
field="ognl:components.inputZip">Zip
</span></th>
<td><input type="text" jwcid="inputZip" size="10"/>
</tr>
<tr>
<td></td>
<td><input type="image" src="images/continue.png"
width="100" height="32"/>
</tr>
</table>
</form>
<hr/>
<p><a href="#" jwcid="@PageLink" page="Home">
Return to Home page</a>.</p>
</body>
</html>
183
184
CHAPTER 5
b
c
d
e
This special ID means that the element will be removed from the template. The
Shell component will provide all of this, and more, in the running application.
This reference to the stylesheet is useful just for WYSIWYG preview.
When using validation, the Form component’s delegate parameter is used to
identify the validation delegate shared by all ValidField components enclosed by
the Form.
Each FieldLabel component is connected to the corresponding ValidField.
Because ValidField components have even more parameters than a TextField component, it is usually best to put all that information in the page specification instead.
As is often the case with Tapestry, the HTML template is just the starting point
for understanding how the page will behave when the application is running.
The following sections fill in the missing details.
Defining a stylesheet for the page
The Register page makes heavy use of CSS; that’s how fields and labels are highlighted in red when they are in error. To support this, the rendered page must
include that stylesheet. You include stylesheets by using a <link> element within
the <head> element (within the <html> element). Because the Shell component is
writing the <html> and <head> elements, that component must write the link to
the stylesheet. This is accomplished by binding its stylesheet parameter to an
asset. A declaration for that asset will appear in the Register page’s specification:
<html jwcid="@Shell" title="Registration"
Removing portions of the template
Now comes a conundrum. We still would like the Register page to preview properly while editing, but if the template includes the <head> and <link> elements
needed to include the stylesheet, then the final rendered page will include two
sets of <head> and <link> elements: one from the template, and one dynamically
rendered by the Shell component.
Tapestry includes a little trick to sidestep this issue:
<head jwcid="$remove$">
<link rel="stylesheet" type="text/css" href="css/style.css"/>
</head>
That special component ID, "$remove$", is the key. It isn’t normally a valid ID
(because it contains a dollar sign). However, Tapestry allows it as a special case
but doesn’t define a new component. Instead, this special ID cues Tapestry to
remove the element and everything enclosed by the element, as if it were never
185
in the template in the first place. With this in place, the page previews correctly
when editing, as demonstrated in figure 5.6.
The Shell component will produce the <html> and <head> element of the rendered page. To support JavaScript within the page, we must use a Body component to render the <body> element.
Figure 5.6
The Register page still previews properly in a WYSIWYG HTML editor.
186
CHAPTER 5
Using a Body component
The ValidField component will almost always produce some client-side JavaScript. At the very least, initialization JavaScript is included that moves the cursor
to the first field that is either required (but empty) or in error. As we discussed in
chapter 4, a Body component is necessary to organize the production of clientside JavaScript. The templates for all examples in this chapter make use of a
Body component:
. . .
</body>
With the basic infrastructure of the page in place, we can move on to the first
dynamic section of the page: the area where any validation error messages
are displayed.
Displaying validation errors
The next section of the template is concerned with displaying validation errors.
The validation delegate for the page is obtained and checked to see if it contains
any errors. The delegate’s hasErrors property will always be false when a page is
initially rendered. The hasErrors property may be set to true during the Form’s
rewind, when a form submission takes place. As you’ll see, the Form’s listener
method will also query the delegate’s hasErrors property. If the property is false,
it is safe to take the validated input and move forward to the next step in the process. If hasErrors is true, the Register page is redisplayed to reveal the errors
and decorate any fields that are in error.
Although it is possible to display all error messages for all fields that are in
error, such output will be unwieldy on large forms and not very useful. Instead,
the error message for the first field that is in error is displayed, but all fields
throughout the form that are in error are marked. You’ve seen examples of this
in figure 5.2, where the message refers to only the first field in error, even
though several fields are marked. The HTML template makes use of Conditional
and Delegator components to produce the formatted error message:
condition="ognl:beans.delegate.hasErrors">
<tr valign="top">
<td>
<img height="52" alt="[Error]" src="images/form-error.png"
width="52"/>
187
</td>
<td>
delegate="ognl:beans.delegate.firstError">
Error Message
</span>
</td>
</tr>
</table>
</span>
In this example, the validation delegate is obtained using the OGNL expression
beans.delegate, a reference to the delegate helper bean (which itself is declared
in the page specification). The Conditional component queries the delegate’s
hasErrors property and displays the error message (formatted inside a <table>)
only if the property is true.
The error objects that the validation delegate returns (such as the firstError
property) are not simply strings but objects that implement the IRender interface. Because these are objects and not strings, we can’t use an Insert component
to display them. Instead, we use a Delegator component, which invokes the
render() method on the renderable object provided to it.
The fact that these are renderable objects, rather than simple strings, opens up a
whole realm of possibilities. The error objects can render all sorts of HTML, not just
text: images, JavaScript pop-up windows, links, formatting—anything HTML. The
intent is that customized validators for the application can provide customized
error objects that are more than just a wrapper around a string. Implementations
that make use of these possibilities are very application specific, which is why the
validator classes provided with the framework don’t use this feature.
The validation delegate, as a convenience, includes a firstError property
that is the renderable object for the first field error. The firstError property is
passed to the Delegator, which results in the error message being displayed.
Following the error message output is the form containing the FieldLabel and
ValidField components.
Starting the Form component
The Form component is used as in examples from the previous two chapters,
with one difference. An additional parameter, delegate, is specified to link the
Form to the validation delegate. Every FieldLabel and ValidField component
enclosed by the Form must use the same validation delegate—and they all will,
since they all will retrieve the validation delegate through the Form:
188
CHAPTER 5
<form jwcid="@Form"
listener="ognl:listeners.formSubmit"
Again, the OGNL expression beans.delegate resolves to a validation delegate.
All the components enclosed by the Form will share this one validation delegate instance.
Using the FieldLabel component
Each ValidField will be preceded in the HTML template by a FieldLabel. The FieldLabel is connected to its partner ValidField by the FieldLabel’s field parameter:
<span jwcid="@FieldLabel"
field="ognl:components.inputFirstName">First Name
</span>
The OGNL expression components.inputFirstName is a reference to the inputFirstName component of the page. Every page provides a read-only Map of all
the components it contains. The keys of this Map are the component IDs. Using
OGNL, you can access the values in a Map just as easily as the properties of an ordinary JavaBean. Of course, to build a reference, you must know the ID of the component, which is one reason the ValidField components are given explicit IDs.
During the render, the FieldLabel discards its body (the text First Name) and
gets, from the ValidField, the correct field name to display as a label. This may
seem cumbersome, but it is useful for two reasons. First, it ensures that the label in
the output HTML matches the name for the field used in any error messages generated by the ValidField’s validator. Second, if the ValidField localizes the name,
the FieldLabel will still match, using the locale-specific value.2 In addition, the validation delegate will have a chance to render before and after the FieldLabel; this
allows the delegate to decorate the label when the field is in error.
Using the ValidField component
In this example, each ValidField is constructed as a declared component, not
an implicit component. Each ValidField appears in the HTML template, but its
type and most of its parameters are declared in the page’s specification. The
ValidField components may not be anonymous—they must have real IDs. This
is necessary so that the placeholder in the template can be linked to the entry
in the page specification, but also so that the FieldLabel can be connected to
the ValidField.
2
Localization of Tapestry applications is discussed in chapter 7.
189
Each ValidField appears in the template minimally, because the bulk of the
component’s configuration is in the page specification:
<input type="text" jwcid="inputLastName" size="50"/>
The FieldLabel and ValidField components blend into the HTML template like
any other type of Tapestry component. They are just a bit more sophisticated in
terms of how they render their HTML (to allow the validators and validation delegate to decorate them) and in how they are configured. When you use Tapestry
form validation, the interesting part takes place inside the page specification.
5.2.2 Creating the Register page specification
The page specification for the Register page (listing 5.3) has two main sections
(beyond the elements you’ve seen before, such as defining the page class and
declaring assets). The first section defines additional helper beans used with this
page, including the validation delegate and the validators used by the ValidField
components. The second section defines each of the ValidField components.
Listing 5.3 Register.page: specification for the Register page
<page-specification class="examples.register.Register">
<bean name="delegate"
class="examples.register.RegisterDelegate"/>
b
lifecycle="page">
</bean>
Declares validation
delegate
c
<bean name="stateValidator"
lifecycle="page">
<set-property name="minimumLength" expression="2"/>
</bean>
<bean name="zipValidator"
class="examples.register.PatternValidator"
lifecycle="page">
<set-property name="pattern">
"\\d{5}(-\\d{4})?"
Declares shared
validator
190
CHAPTER 5
</set-property>
<set-property name="errorMessage">
"Zip code format is five or nine digits.
Example: 02134 or 02474-1145."
</set-property>
</bean>
<property-specification name="address"
type="examples.register.Address"/>
<component id="inputFirstName" type="ValidField">
<static-binding name="displayName"
value="First Name"/>
<binding name="validator"
expression="beans.required"/>
<binding name="value"
expression="address.firstName"/>
</component>
d
Declares
ValidField
component
<component id="inputLastName" type="ValidField">
<static-binding name="displayName" value="Last Name"/>
<binding name="validator" expression="beans.required"/>
<binding name="value" expression="address.lastName"/>
</component>
<component id="inputAddress1" type="ValidField">
<static-binding name="displayName" value="Address"/>
<binding name="value" expression="address.address1"/>
</component>
<component id="inputCity" type="ValidField">
<static-binding name="displayName" value="City"/>
<binding name="value" expression="address.city"/>
</component>
<component id="inputState" type="ValidField">
<static-binding name="displayName" value="State"/>
<binding name="validator" expression="beans.stateValidator"/>
<binding name="value" expression="address.state"/>
</component>
<component id="inputZip" type="ValidField">
<static-binding name="displayName" value="Zip"/>
<binding name="validator" expression="beans.zipValidator"/>
<binding name="value" expression="address.zip"/>
</component>
<context-asset name="stylesheet" path="css/style.css"/>
b
c
d
191
The validation delegate is defined here, using a custom subclass, examples.register.RegisterDelegate. It can be referenced using the OGNL expression
beans.delegate. The default lifecycle, “request”, means that the bean will be discarded at the end of the current request.
Most of the fields share a single instance of StringValidator as the required
bean. Since validators don’t store any internal state, the “page” lifecycle means
that the bean, once created, will be kept for as long as the page instance itself.3
Each ValidField follows this same pattern, providing a displayName (used by the
FieldLabel and in any error messages), a validator (which may be shared
between different fields), and a value to edit.
This specification includes yet another new element, <static-binding>. The
<static-binding> element is used to set a component property to a fixed, literal
string. Such a string is static (unchanging), in contrast to the dynamically evaluated OGNL expression used in a <binding> element. The <static-binding> element is used throughout the specification to provide the displayName parameter
for each ValidField.
As an alternative, the value could appear in the HTML template:
<input type="text" jwcid="inputFirstName" size="50"
displayName="First Name"/>
Either way, the displayName parameter will be set to a static string value. Tapestry
is completely flexible as to where parameters are bound (in the specification or
in the HTML template).
TIP
A good standard is to define formal parameters only in the page specification and informal parameters only in the HTML template. This
scheme reflects that formal parameters are most often behavioral and
informal parameters are more often related to presentation. Using this
division keeps you from wasting time tracking down the right file when
a change is needed.
This specification declares several helper beans and a number of ValidField components. The first bean defined is the validation delegate.
Defining the validation delegate and validators
The RegisterDelegate class is an implementation of the IValidationDelegate
interface that customizes the rendering hooks of the delegate to reference CSS
3
Remember that page instances, once constructed, are cached for later reuse in subsequent requests.
192
CHAPTER 5
styles and images specific to the Register application. We’ll examine its code in
detail shortly.
The majority of the fields in the form are used to edit simple string properties, and the only validation constraint applied to them is that a non-null value
be supplied by the user. All such ValidField components will share a bean named
required as their validator.
The required bean instantiates an instance of StringValidator and sets the
required property of the StringValidator to true. This will force users to provide
a non-empty value for the field. Like most validators, the StringValidator has
no internal state and can therefore be retained indefinitely once created (which
is what the page lifecycle does).
The delegate bean should not be given a lifecycle of page because it has a considerable amount of internal state (the error messages for fields that are in error)
that is relevant only to a single request from a single user.
The ValidField for the state property uses a different validator, stateValidator.
This validator is also declared as a helper bean in the page specification:
lifecycle="page">
</bean>
Like the required bean, the stateValidator bean sets the required property to
true, forcing the user to provide a value. It adds a second constraint, configured
through a second property, requiring the input to be at least two characters in
length. The normal size attribute (specified as an informal parameter in the
HTML template) ensures that no more than two characters are provided.
The final validator is for the zip code field. Zip codes have a pattern that is
best described using a regular expression. Tapestry doesn’t provide a validator
along these lines, but it is easy enough to create one. Although we could create a
validator that supports only the zip code pattern, it is virtually no extra work to
create a flexible validator where we can configure both the regular expression
pattern we want to validate against as well as the error message we want to display if the input fails to match the pattern.
In terms of maximizing usability, it is important that error messages be as
helpful to the user as possible. Simply telling users that their input didn’t match
a regular expression would leave them frustrated and at a loss as to how to correct their input to satisfy the application. Instead, you can display a custom error
193
message that tells users exactly what they need to do. For the zip code field, we
can customize the error message to give users examples of the two zip code formats accepted. The regular expression pattern and the error message are both
provided in the page specification:
<bean name="zipValidator"
class="examples.register.PatternValidator"
lifecycle="page">
<set-property name="pattern">
"\\d{5}(-\\d{4})?"
</set-property>
<set-property name="errorMessage">
"Zip code format is five or nine digits.
Example: 02134 or 02474-1145."
</set-property>
</bean>
Here we are employing an alternate usage of the <set-property> element.
Instead of specifying an expression attribute, as in the previous examples, we are
putting the OGNL expression in the body of the <set-property> element. This
approach is useful here, because we must enclose the literal string values (the
pattern and the error message4) in double quotes. Putting the OGNL expression
in the body is much easier when the expression is long or contains complex
punctuation, such as a mix of single and double quotes.
Now that we have declared the validation delegate and the three validators,
we can continue on to the ValidFields themselves.
Declaring the ValidField components
All the ValidField component declarations follow the same general template. A
displayName for the field is provided; this is used by the FieldLabel and in any
error messages created by the field’s validator. Like an ordinary TextField component, the value parameter is bound to the property the ValidField will edit.
Unlike a TextField component, however, this property can be bound to any type
of property, not just string properties. Finally, a validator is specified. The validator defines the type of input acceptable in the field; similar fields can share validators. The first ValidField is used for entering the user’s first name:
<component id="inputFirstName" type="ValidField">
<static-binding name="displayName" value="First Name"/>
4
Alternately, we could put the message into a string properties file, and use the <set-message-property>
element to retrieve the localized message and set the bean property from it. Appendix D has a complete description of the specification DTD.
194
CHAPTER 5
<binding name="value" expression="address.firstName"/>
</component>
The inputFirstName component uses the required bean as its validator and edits
the firstName property of the address.
All the remaining ValidField components follow the same pattern, providing different display names, different properties, and different page properties for editing.
Implementing the Register page
The Java code for the Register page is concerned with providing the address
property referenced by the many ValidField components as well as handling the
form submission. Listing 5.4 shows the Register class. We use the same lifecycle
technique as before to initialize the address property to a non-null value before
the page initially renders (and when the form is submitted). We must check for
null because if the page renders again (following a form submission with an
input validation error), the pageBeginRender() method will be invoked again
(first because of the form rewind, and then again for the page render).
Listing 5.4 Register.java: Java class for the Register page
import
import
import
import
import
org.apache.tapestry.event.PageRenderListener;
org.apache.tapestry.valid.IValidationDelegate;
public abstract class Register extends BasePage
implements PageRenderListener
{
public abstract Address getAddress();
public abstract void setAddress(Address address);
{
if (getAddress() == null)
setAddress(new Address());
}
{
IValidationDelegate delegate =
(IValidationDelegate) getBeans().
getBean("delegate");
c
b
Initializes the
address property
Accesses the
delegate bean
Validating input based on regular expressions
195
if (delegate.getHasErrors())
return;
RegisterConfirm next =
(RegisterConfirm) cycle.getPage("RegisterConfirm");
next.setAddress(getAddress());
d
Passes the Address
to the next page
}
}
b
c
d
This method will be invoked when the page initially renders, when the form
within the page is submitted, and again if the page is rendered after the form is
submitted (because of validation errors). The ValidField components require that
a non-null Address be available in the address property.
In Java code, accessing helper beans involves the getBeans() method, from
which individual beans can be retrieved by name.
Passing along information collected in a form to another page, shown here, is a
standard technique.
The listener method, formSubmit(), is invoked when the Form is submitted. In
order to determine if there were any errors, the code accesses the validation delegate. If there are errors, the method returns, causing the Register page to redisplay with the error message shown and invalid fields highlighted.
If there are no errors, then the listener method activates the next page, which
shows a confirmation. A real application would save this address information to a
database before continuing, but that is beyond the scope of this example.
Most of this example is implemented using standard components and
objects, but to properly validate the input provided in the zip code field, we need
a custom validator class.
5.3 Validating input based on regular expressions
Sometimes, you will need to perform validations for which there is no out-of-thebox validator. In the Register example, validating that a user’s zip code is properly formatted falls into that category. There’s a specific regular expression that
can be used, which checks if the input is a traditional five-digit zip code or an
extended nine-digit zip code.
The framework doesn’t include a validator that can employ regular expressions;
but creating validators in Tapestry involves just two methods, so it’s simple to create
one of our own. Listing 5.5 is the implementation of this validator.
196
CHAPTER 5
Listing 5.5 PatternValidator.java: validator based on regular expressions
import
import
import
import
import
import
import
import
import
org.apache.tapestry.form.IFormComponent;
org.apache.tapestry.valid.BaseValidator;
org.apache.tapestry.valid.ValidatorException;
org.apache.oro.text.regex.MalformedPatternException;
org.apache.oro.text.regex.Pattern;
org.apache.oro.text.regex.PatternCompiler;
org.apache.oro.text.regex.Perl5Compiler;
org.apache.oro.text.regex.Perl5Matcher;
public class PatternValidator extends BaseValidator
{
private String _pattern;
private Pattern _compiledPattern;
private String _errorMessage;
private Perl5Matcher _matcher;
public String toString(IFormComponent field,
Object value)
{
if (value == null)
return null;
b
Converts object
to string
return value.toString();
}
public Object toObject(IFormComponent field,
String input)
throws ValidatorException
{
if (checkRequired(field, input))
Returns if
return null;
c
d
Converts
submitted value
input blank
if (!match(input))
throw new ValidatorException(errorMessage, null);
e
return input;
}
protected boolean match(String input)
{
if (_compiledPattern == null)
{
PatternCompiler compiler = new Perl5Compiler();
f
try
{
Matches
input
against
regular
expression
Compiles
and caches
regular
expression
Validating input based on regular expressions
_compiledPattern = compiler.compile(_pattern);
}
catch (MalformedPatternException ex)
{
throw new ApplicationRuntimeException(ex);
}
197
f
}
if (_matcher == null)
_matcher = new Perl5Matcher();
return _matcher.matches(input, _compiledPattern);
}
public String getErrorMessage()
{
return _errorMessage;
}
public String getPattern()
{
return _pattern;
}
public void setErrorMessage(String errorMessage)
{
_errorMessage = errorMessage;
}
public void setPattern(String pattern)
{
_pattern = pattern;
_compiledPattern = null;
Updates pattern
}
and clears cache
g
}
b
c
d
e
The toString() method converts the value (obtained from the ValidField’s
value parameter) into a string, which becomes the value attribute of the
<input> element.
The toObject() method converts a string entered by the user back into an object
value, throwing a ValidatorException if the input is improperly formed or otherwise invalid.
The checkRequired() method checks to see if the input is blank. It throws a
ValidatorException if the input is blank and the field is required. Otherwise, it
returns true if the input is blank and false otherwise.
The match() method does the regular expression pattern matching.
198
CHAPTER 5
f
g
Converting a string into a regular expression is somewhat expensive, so it’s done
only the first time match() is invoked.
If the pattern changes, the cached compiled pattern should be discarded.
PatternValidator extends the framework class BaseValidator. BaseValidator is
abstract and implements the IValidator interface. The BaseValidator class provides the boolean required property, plus a bit of support for client-side script-
ing (which allows the validator to generate client-side JavaScript to perform
validations entirely within the client web browser).
The two key methods in a validator are toString() and toObject(). The
toString() method is used to convert an object (read from the ValidField’s value
parameter) into a string. This method is used when the ValidField renders; the
converted string is used as the value attribute of the HTML <input> element.
A validator should always be able to translate a null value to a string. It is
acceptable to return null from toString() if the value passed in is null; this is
what the validators provided with the framework do.
The meat of the validator is in the complementary toObject() method. This
method is invoked when the form is submitted. The purpose of toObject() is to
convert a string, supplied by the end user, into an object, such as an Integer or
Date—whatever is appropriate for the specific type of validator. This is where all
conversions and validations take place. If the string can’t be converted, or the
value is invalid for other reasons, the method throws a ValidatorException that
is caught by the ValidField and used to record an error for the field.
The first step in the PatternValidator’s implementation of toObject() is to
invoke the method checkRequired(), which is supplied by the BaseValidator
class. This method performs two functions: It returns true if the input is null
or empty (an empty string is length zero, or contains only whitespace), and it
also throws a ValidatorException if the validator is required (and the input is
null or empty). As configured in the Register page, the zip code field is
optional (not required), so if the user decides not to enter a value, the field will
not be in error.
Assuming a value was supplied by the user, the toObject() method continues
by invoking the match() method. If the input from the user does not match the
regular expression pattern configured for the validator, then a ValidatorException
is thrown. The exception is built around the supplied error message (another
configurable property of the PatternValidator class). This exception is caught
by the validation delegate, which records the error message for later use when
the page is rendered again (to display the errors to the user).
Customizing label and field decorations
199
If match() returns true, then the input value becomes the return value for the
toObject() method. Ultimately, this value will be assigned to the property bound
to the ValidField’s value parameter.
The match() method uses the Jakarta ORO framework to compile the pattern
and match the compiled pattern against user input. The PatternValidator does
a little caching, since compiling a string to a Pattern object is somewhat expensive; it shouldn’t be done every time.
5.4 Customizing label and field decorations
In figure 5.1, all the required fields are marked with a red asterisk (to the right of
the field). A glance at the HTML template for the page, in listing 5.2, shows that
these markers are not in the template itself. This is an example of field decoration, one of the functions of the validation delegate. Additionally, you’ve seen
that FieldLabels and ValidFields are also decorated when they are in error.
The base implementation of the IValidationDelegate interface, ValidationDelegate, provides all the support for tracking fields and errors, as well as simple
support for decorating fields and labels. To customize the look and feel, as
shown in figure 5.2, you create your own subclass, overriding several methods
related to field and label decoration.
For the Register page, just such a subclass of ValidationDelegate is shown in
listing 5.6. It overrides several methods supplied in the ValidationDelegate base
class, supplying application-specific look and feel.
Listing 5.6 RegisterDelegate.java: validation delegate subclass for the Register page
import
import
import
import
import
org.apache.tapestry.valid.IValidator;
org.apache.tapestry.valid.ValidationDelegate;
public class RegisterDelegate
extends ValidationDelegate
{
public void writeLabelPrefix(
IFormComponent component,
IMarkupWriter writer,
{
b
Subclasses from
ValidationDelegate
Is called before
FieldLabel
renders
200
CHAPTER 5
if (isInError(component))
{
writer.begin("span");
writer.attribute("class", "label-error");
}
c
Encloses label
with <span>
d
Is called as
ValidField
renders
}
public void writeLabelSuffix(
{
if (isInError(component))
{
writer.end(); // span
}
}
Is called after
FieldLabel
renders
public void writeAttributes(
IRequestCycle cycle,
IValidator validator)
{
if (isInError())
writer.attribute("class", "field-error");
}
public void writeSuffix(
IValidator validator)
{
if (validator != null &&
validator.isRequired())
{
writer.printRaw(" ");
writer.begin("span");
writer.attribute("class",
"required-marker");
writer.print("*");
writer.end();
}
if (isInError())
{
writer.printRaw(" ");
writer.beginEmpty("img");
writer.attribute("src",
"images/field-error.png");
e
Is called after
ValidField renders
Marks
required
fields
Marks fields
that are in
error
Customizing label and field decorations
writer.attribute("width", 16);
writer.attribute("height", 16);
}
}
201
Marks fields
that are in
error
}
b
c
d
e
The framework class ValidationDelegate provides the majority of the behavior
for a validation delegate. Generally, a subclass is needed only as in this example:
to override the methods used for decorating fields and labels.
If the field for the label is in error, then an HTML <span> is wrapped around it,
with a class of label-error. The stylesheet for the page modifies the enclosed
text to make it red.
writeAttributes() is invoked as the ValidField writes the <input> tag and allows
the validation delegate to write additional attributes if the field is in error. Here,
the CSS class field-error is written into the <input> tag if the field is in error,
resulting in the white-on-red display.
writeSuffix() is invoked after the ValidField renders the <input> tag. The
IValidationDelegate interface defines an additional method, writePrefix(),
which is invoked before the ValidField renders; and the ValidationDelegate base
class provides a do-nothing implementation of that method.
The first two methods in the class, writeLabelPrefix() and writeLabelSuffix(),
are related to decoration of field labels. The FieldLabel component invokes these
methods, passing the ValidField it is connected to as the component parameter.
These two methods rely on the isInError() method, which returns true if a validation error has been recorded for a component. When isInError() returns true,
the delegate renders a <span> tag and writes the class attribute for it (as labelerror). The IMarkupWriter interface is provided by Tapestry; it is much like a
java.io.PrintWriter, but with additional methods for streamlining the output
of elements and attributes, as shown in listing 5.6. Chapter 7 includes more
information about IMarkupWriter.
The <span> tag started in writeLabelPrefix() is ended inside the writeLabelSuffix() method when it invokes end() on the writer. In between the two
methods, the FieldLabel gets the displayName from the ValidField and writes that
to the writer. For example, if the First Name field is in error (because the user
submitted the form without filling in a value), then the FieldLabel, combined
with the validation delegate, will render the following:
<span class="label-error">First Name</span>
202
CHAPTER 5
Just as the FieldLabel delegates part of its rendering to the validation delegate,
so does the ValidField. Three methods of the validation delegate are used to decorate fields: writePrefix(), writeSuffix(), and writeAttributes(). RegisterDelegate implements the latter two methods; the default writePrefix() method,
inherited from the ValidationDelegate base class, does nothing. The writePrefix() method is invoked by the ValidField before it renders the <input> tag.
The writeAttributes() method is called by the ValidField after it has written
the <input> tag and all of its attributes, but before it closes the tag. This gives the
validation delegate a chance to add attributes to the tag. In this case, the delegate checks if the current component is in error; if so, a class attribute is added
to the <input> tag. This CSS class is combined with the page’s stylesheet to display offending fields as white text against a red background.
The writeSuffix() method is invoked by the ValidField after it closes the
<input> tag. RegisterDelegate queries the ValidField’s validator (which is passed
in to the method as a parameter) to see if the validator requires a value. If so, the
delegate writes HTML to display the required marker. Likewise, if the field is in
error, HTML is written that marks the field using a specific image.
NOTE
Normally, it is not necessary for your implementation of the writeSuffix()
method to check that the validator parameter is not null. A ValidField
component will always have a validator and will pass it into this method.
This check is necessary because, in chapter 8, we create a new type of
component that uses a validation delegate but does not have a validator.
As you can see, the methods defined by the IValidationDelegate interface are
designed to be completely open-ended. With a minimal amount of effort, you
can customize the look and feel of your application’s labels and fields by creating
and using your own subclass of the ValidationDelegate base class.
So far, we’ve covered how the validation subsystem extends normal form processing (both when rendering a page and when the form is submitted). That’s
only part of the story, however, since the validation framework can also contribute to the client side.
5.5 Enabling client-side validation
Performing validations when the form is submitted is a powerful approach, but
an even better solution is to not submit the form until the fields are valid within
the client web browser. Using a client-side solution gives users more immediate
feedback (as soon as they click the submit button) because no additional round-trip
to the server is required. Server-side validation will still occur when the form is
Enabling client-side validation
203
submitted, since the client can’t be counted on to have JavaScript enabled. In
addition, for basic security and data integrity reasons, you must never rely on the
client web browser to enforce any rules (in fact, you can’t even count on the software at the other end of the HTTP connection to be a web browser; it could be a
web spider, a screen scraper, or some kind of script).
Using Tapestry and the ValidField component, you can integrate this kind of
client-side validation seamlessly into your applications. The ValidField’s validator
is responsible for generating this client-side JavaScript. When enabled, JavaScript
event handlers are created and hooked into the form submission to perform validations within the client web browser similar to those that occur in the server.
The examples include a second version of the Register page, Register2. It is
virtually identical to the first page, except that in the page specification, client
scripting is enabled. If users attempt to submit the form with missing or invalid
data, a client-side pop-up window advises them of the error. Figure 5.7 demon-
Figure 5.7 With client validation enabled, submitting the form causes validations to occur.
After the user clicks the OK button, the cursor will be moved to the Last Name field.
204
CHAPTER 5
strates what happens if you submit the form without providing a value for the
Last Name field, which is required.
Enabling this support requires a small change to the page specification.
The bean property clientScriptingEnabled must be set to true for each of
the validators:
lifecycle="page">
<set-property name="clientScriptingEnabled" expression="true"/>
</bean>
lifecycle="page">
<set-property name="clientScriptingEnabled" expression="true"/>
</bean>
To support browsers where JavaScript does not exist or is disabled, all validations still occur on the server when the form is submitted. Some validations may
not be possible on the client side; they may be too complex to express in JavaScript or require access to data that isn’t available in the client, such as information from a database.
You’ve already seen how to create new validators that perform server-side validations. Adding client-side validations for a new validator is more involved; Tapestry includes the necessary tools for dynamically generating the JavaScript (this
is covered in chapter 8), and the IValidator interface includes a method for this
purpose, renderValidatorContribution(). Nevertheless, getting validations to
work still requires a fairly deep understanding of both Tapestry and JavaScript.
To accomplish client-side validation, each validator must
■
Create a JavaScript function that accesses the field value and performs the
validations; if invalid, the function must display an error window and
return false
■
Adapt the function to Tapestry-generated IDs for the form and the text field
■
Register the script function with the Form component as a client-side submit event handler
The BaseValidator class, a base class implementing IValidator from which most
validator classes extend, includes several methods for supporting client-side
205
scripting. The key challenge is the correct generation of the JavaScript needed
in the client, which uses techniques discussed in chapter 8.
In addition to adding client-side scripting support to new validators, you can
change the client-side scripting support for the existing validators. The existing
validators provide reasonable client-side scripting support. As shown in figure 5.7,
invalid input will cause a pop-up window to appear that indicates the problem
and names the field; clicking the OK button will set input focus to the field in
error and select all text in the field. A demanding application may want to do
more—for example, change the CSS class for fields that are in error to visually
highlight such fields for the user, or display the error messages for fields on the
page itself, rather than in a pop-up window. Such application-specific functionality will require application-specific client-side JavaScript, which can be supplied
by configuring the validators’ instances with the custom scripts (again, using
techniques described in chapter 8).
So far, all discussion of validation has concerned individual fields in isolation.
In many cases, there are dependencies between fields that must also be validated—form-level validations that involve two or more fields.
5.6 Handling form-level validations
Validations that involve more than one field occur inside the Form’s listener
method. For example, a form-level validation may check that two input dates are
in ascending order, and display an error message if they are not. This is demonstrated in figure 5.8.
There’s no need to throw away the rest of the validation subsystem to handle
these kinds of cases; it’s possible for a Form’s listener method to mark fields in
error, just as easily as a validator can. This is accomplished by putting additional
validation logic in the Form’s listener method.
The example shown in figure 5.8 is for the Dates page. This page uses two ValidField components, inputStart and inputEnd, to edit two page properties (startDate
and endDate, respectively). Listing 5.7 shows the page class for the Dates page.
Listing 5.7 Dates.java: Java class for the Dates page
package examples.dates;
import java.util.Date;
import
import
import
import
206
CHAPTER 5
Figure 5.8 Although the input values are valid dates, the end date precedes the start date and is
rejected with an error message.
import org.apache.tapestry.valid.ValidationConstraint;
public abstract class Dates extends BasePage
{
public abstract Date getStartDate();
public abstract Date getEndDate();
{
Obtains validation
(IValidationDelegate) getBeans().
delegate helper bean
getBean("delegate");
return;
Returns if ordinary
validation errors occurred
207
Date startDate = getStartDate();
Date endDate = getEndDate();
if (startDate.after(endDate))
{
IFormComponent inputEnd =
(IFormComponent) getComponent("inputEnd");
delegate.setFormComponent(inputEnd);
delegate.record(
"End Date must be after Start Date.",
ValidationConstraint.CONSISTENCY);
return;
}
b
Records a
form-level
validation
error
DatesConfirm next =
(DatesConfirm)cycle.getPage("DatesConfirm");
next.setStartDate(startDate);
next.setEndDate(endDate);
}
}
b
The error is attributed to the second ValidField, the one that gets the end date.
The call to setFormComponent() identifies the field to be assigned the error. The
call to record() provides the error message.
This listener method starts in much the same way as in the previous example;
it gets the validation delegate and simply returns if the delegate already has
errors. Validation errors at this point are formatting errors in the user input,
or are the result of the user omitting one of the fields (both of which are
required). The remaining code in the method is executed only if there are no
fundamental input errors, in which case both the startDate and endDate properties will have been supplied by the respective ValidField components (inputStart and inputEnd).
If the start date occurs after the end date, then a form-level validation error
occurs. The first step is to identify the field to be associated with the error. This
isn’t strictly necessary; it is acceptable to not invoke the setFormComponent()
method, in which case the error is recorded but will not be associated with any
specific field within the form.
In this example, we identify the end date as the error field (we could just as
easily mark the start date). The validation delegate’s record() method lets us
assign an error to the field. When the page is re-rendered, the inputEnd field
will be marked in error. There are several implementations of the record()
208
CHAPTER 5
method, each taking different parameter types for the message. The final parameter is of type ValidationConstraint, an Enum of different possible reasons the
field is in error.5 If none of the provided constraint values are appropriate, then
null is an acceptable value. CONSISTENCY is used to indicate that a cross-field consistency error occurred.
Once the field is recorded as containing an error, the listener method returns.
This action will cause a redisplay of the active page, complete with error messages and field decorations.
So far, you’ve seen a number of ways to mix and match FieldLabel and ValidField components with validators and validation logic inside listener methods.
With some extra effort, it is possible to use much of the validation system without
using the ValidField component, as you’ll see in the next section.
5.7 Using validation without ValidField
There are times when you will want to mix and match, leveraging portions of
the validation subsystem without necessarily using the ValidField component.
For example, figure 5.9 shows a different version of the page from the previous
example. This page uses DatePicker components, instead of ValidField components, to collect the start and end date from the user, yet it still uses the rest of
the validation framework (FieldLabel components error display, and a validation delegate).
For the most part, the HTML template for the revised Dates page is the same
as in the prior example (using ValidField), with some minor differences around
the labels and fields. It is possible to use the FieldLabel component, although
the HTML template is somewhat more involved:
<tr>
displayName="Start Date"
field="ognl:components.inputStart">
Start Date</span></th>
<td><input type="text" jwcid="inputStart"
size="10"/>
5
b
Labels the field
as before
c
Contains the DatePicker
component
The validation constraint is not used by the default implementation of IValidationDelegate or by
our custom subclass. It is provided in speculation that a more clever validation delegate implementation may have a need for it—such as customizing the decoration for an invalid field based on the type
of constraint violated.
Using validation without ValidField
Figure 5.9
The DatePicker component can be used with the form validation framework.
<span class="required-marker">*</span>
condition="ognl:beans.delegate.inError">
<img src="images/field-error.png"/>
</span>
</td>
</tr>
b
c
209
Decorates
the field
The FieldLabel component can still be used, but the display name must be specified as a parameter.
Without a ValidField, the validation delegate is not integrated into the render.
The template must include the decorations normally provided by the validation
delegate: the marker for required fields and, optionally, the image used to mark
error fields. The validation delegate’s inError property is true when the current
field is marked as in error.
When using a ValidField, the displayName (used by the FieldLabel component) is
provided as a parameter of the ValidField. Without a ValidField, you must instead
set the displayName parameter directly on the FieldLabel component.
210
CHAPTER 5
All form-related components within a Form are tracked by the validation delegate even if they are not ValidField components. This allows errors to be
attached to any kind of form control component: TextField, TextArea, PropertySelection, even DatePicker (as in this example). However, the validation delegate
is not integrated into the rendering of these other types of components; only
ValidField has all the necessary hooks. The work the validation delegate normally does, decorating invalid fields, is instead done explicitly in the template,
using a Conditional component.
The validation delegate’s inError property is true if the most recently rendered component has an error. Because the Conditional in the template follows
the DatePicker component, the inError property will be true if there was a validation error for the DatePicker.
Because there is no ValidField, and therefore no validator, involved in form
processing, even simple checks for required fields must now be done inside the
Form’s listener method:
{
(IValidationDelegate) getBeans().getBean("delegate");
Date startDate = getStartDate();
Date endDate = getEndDate();
if (startDate == null)
error(delegate, "inputStart",
"Start Date is required.",
ValidationConstraint.REQUIRED);
Gets values set by
DatePicker components
b
if (endDate == null)
error(delegate, "inputEnd",
"End Date is required.",
ValidationConstraint.REQUIRED);
Checks for
missing
required fields
return;
if (startDate.after(endDate))
{
error(
delegate,
"inputEnd",
"End Date must be after Start Date.",
ValidationConstraint.CONSISTENCY);
return;
}
c
Performs
form-level
check
Summary
211
DatesConfirm next =
(DatesConfirm) cycle.getPage("DatesConfirm");
next.setStartDate(startDate);
next.setEndDate(endDate);
}
private void error(
IValidationDelegate delegate,
String componentId,
String message,
ValidationConstraint constraint)
{
IFormComponent component =
(IFormComponent) getComponent(componentId);
delegate.setFormComponent(component);
delegate.record(message, constraint);
}
b
c
These checks are normally done by the validator for the ValidField. Since the
form uses DatePicker components instead of ValidField components, there is no
validator to perform even this basic check.
This form-level check is the same as in the previous example (which used ValidField components).
Despite the larger amount of code (some of which could be factored out to base
classes or helper beans), the process is still straightforward: We check the properties edited by the DatePicker components, compare the values, and inform the
validation delegate if there is an error.
5.8 Summary
Tapestry’s validation subsystem, centered around the ValidField component, is a
tremendous boon to web application usability. The validation subsystem achieves
the framework’s goals for simplicity and consistency, both from the end user’s
perspective and from your perspective as the application developer. Tapestry
comes with a number of predefined validations but is completely open-ended in
terms of adding new ones. With only a small amount of coding, it is possible to
precisely control the look and feel of label and field decorations.
Once again, large amounts of coding disappear into the framework, and what
little coding remains (in terms of new validators and validation delegates) can be
easily reused within a single application, or even across multiple applications.
Tapestry makes it simple to create a polished, usable user interface.
212
CHAPTER 5
Along the way, you’ve also seen another technique for extending the functionality of the application with little or no coding: helper beans. Helper beans fit
into the overall Tapestry framework of objects, methods, and properties, providing another dynamic way to glue custom behavior into prepackaged components
such as the ValidField.
In addition, you’ve seen that the validation subsystem can play well with
ordinary form control components, such as the DatePicker component. You are
free to mix and match the pieces that best solve your particular application’s
issues. The ordinary form control components, the advanced components such
as PropertySelection and DatePicker, and the validation subsystem can all be
used together to provide a comprehensive, yet still light and agile, solution to
handling all variations of form input in a Tapestry application.
Part 2
Creating Tapestry components
C
hapters 6, 7, and 8 take you beyond using existing components and
show you how to create your own components. Along the way, you’ll see more
of Tapestry’s internals, including how Tapestry applications can be localized.
You’ll also learn how Tapestry components can generate client-side JavaScript, and how Tapestry and traditional servlet applications can interoperate.
Creating reusable
components
This chapter covers:
■
Creating component templates
and specifications
■
Understanding component parameters
■
Working with components packaged in libraries
■
Creating component libraries
215
216
CHAPTER 6
So far in this book, you’ve seen several interesting ways of combining the components provided with the Tapestry framework to form pages, but the real fun
starts when you begin creating brand-new components. Creating components
with Tapestry is easy—in fact, you’ve been doing it all along, since pages are just
a specialized type of component.
Components are the key to high levels of reuse, the Holy Grail of objectoriented development. Creating customized components for your application
and then reusing them on different pages of your application, or even in entirely
different applications, is a massive boost to productivity for you and your developers, because it reduces the amount of Java code and HTML templates you will
have to write. It also improves the robustness of your application by cutting down
on testing; you use the same component in different places, rather than duplicating templates and functionality. Remember, less code always translates to
fewer bugs!
In this chapter, you’ll learn how to create components that use their own
templates and how to create components that forgo a template and produce
output exclusively in Java code. You’ll also see how to use components that
come in a library, and how to create and package a component library for use in
new applications.
6.1 Creating simple template components
The easiest components to create are those designed for a specific use within a
specific application. This type of component may be used in many pages of a single application but is not applicable to other applications and, very often, exploits
specific knowledge about the page or application for which it was created.
Fully reusable components must be completely free of any dependencies on a
specific application. They rely heavily on the use of parameters that configure
and adapt them for a particular usage within a certain application.
To some degree, an application-specific component is like a compiler macro
(a concept familiar to C programmers). Simple components with templates
expand when the page is loaded, adding their template and the components
within it to the containing page’s template.
By way of an example, we’ll revisit the Hangman application from chapter 2.
If you remember, it contained a lot of duplicated code related to converting
numbers and characters into images (to support the “chalkboard” look and feel).
Specifically, the Guess, Win, and Lose pages contained duplication, both in the
page classes as well as their page specifications. Particularly onerous was the
Creating simple template components
217
long, long list of context assets that defined the images for each letter, for each
digit, and for each version of the hanged man in the gallows. Different portions
of this list were duplicated in the Guess, Win, and Lose page specifications.
An obvious first step is to isolate all that image manipulation into a few reusable components, with the goal that each image asset be defined only once,
within a single component specification. These components will be an example
of application-specific reuse; that is, they will be completely tied to this specific
application. We’ll create components for displaying the numbers and the scaffold
used to show the player’s remaining guesses, as well as a component for displaying a particular letter. We’ll then build on these, creating a component that spells
out entire words. You’ll also see a common type of component, known as a border
component, which is used to provide the specific look and feel of an application
(in terms of Cascading Style Sheets [CSS], basic navigation controls, and tablebased layout). These are all examples of the most common types of components
you’ll be creating in your own applications; these components save us a bit of
effort in this application, even if they won’t help when we code our next one.
When we’re done, the revised Hangman application will look and feel exactly
the same as the original; only its internal structure will be different. When components are created and utilized properly, the result is smaller, simpler HTML
templates and page specifications, less code, and, more important, less duplication of code—always a lifesaver when it comes time to fix bugs.
The source code for both the original and the revised Hangman applications
is available online; appendix B discusses how to obtain the source, as well as how
to build and deploy all the examples in the book. The improved Hangman
described in this chapter is accessible as http://localhost:8080/hangman2/app.
Creating a template component entails creating the component specification
and, optionally, the component’s template, and usually involves creating a new
Java class. All three together form a component. Figure 6.1 shows these artifacts.
The component specification is stored in the WEB-INF folder and has a .jwc
extension. The Java class is identified in the specification (defaulting to BaseComponent if left unspecified).
When a page template or specification references a component, Tapestry
must determine how to instantiate an instance of the component’s class, as a part
of the overall page. The first step is to locate the component’s specification,
which (like a page specification) identifies the Java class Tapestry will instantiate
(along with many other implementation details, such as additional embedded
components). Also, most components will have their own HTML template, whose
contents are integrated into the containing page’s template.
218
CHAPTER 6
Figure 6.1 Like pages, components have a specification (with a .jwc
extension) and a Java class. If they have a template, it is stored in WEB-INF
with the specification.
Tapestry starts with the component type, which is used as the name of the component specification file that you want to read. These files may be built-in components bundled with the Tapestry framework, or they may come from a
component library (discussed in section 6.8), or they may be application-specific
components, with the specification stored in the WEB-INF folder.
After reading the specification and instantiating and configuring an instance
of the component’s Java class, Tapestry will then locate and parse the HTML
template for the component.1 The template for the component is stored in the
WEB-INF folder, alongside the component specification (a marked difference
from pages, where the template is stored one level up, in the root context
folder). A template will be used only with components that inherit from BaseComponent—components that subclass from AbstractComponent work entirely in
Java code, with no template.
Regardless of whether the component has a template, it must have a component specification.
1
In fact, the component itself is responsible for locating its template; this is covered in detail in
chapter 7.
219
6.2 Creating the component specification
A component specification tells Tapestry how to instantiate and use a component, in exactly the same way that a page specification tells Tapestry how to
instantiate and use a page. A component specification includes all the same
attributes and elements as a page specification:
■
■
■
■
A class attribute to identify the Java class you want to instantiate
<property-specification> elements to define additional properties for
the component
<component> elements to declare additional components contained within
the component
<context-asset> elements to declare assets used by the component
What a component specification adds (beyond what is available in a page specification) are additional attributes and elements to define if and how the component uses parameters, and how it fits into the overall page:
■
■
■
■
<parameter> elements declare each formal parameter, providing a name
and type, and indicating if the parameter is required or optional.
<reserved-parameter> elements filter out unwanted informal parameters.
The allow-informal-parameters attribute specifies whether informal parameters are allowed or discarded.
The allow-body attribute specifies whether the body of the component (in
its container’s template) is kept and used, or discarded.
This chapter covers the most common aspects of the framework’s component
specification. Appendix D contains a complete reference to all of the Tapestry specifications.
6.2.1 Specifying the component’s Java class
A component specification begins with a <component-specification> element. As
with a page specification, the <component-specification> element includes a
class attribute, which is used to specify which Java class to instantiate for the component. The attribute is optional, in which case the default class, BaseComponent, is
used. BaseComponent is the base class for components that have a template.2
2
BaseComponent’s superclass is AbstractComponent. AbstractComponent is for components that don’t
use a template. Subclasses of AbstractComponent must implement the renderComponent() method
to produce their output entirely in code.
220
CHAPTER 6
When creating new components, it is often (but not always) necessary to create a custom subclass of BaseComponent to contain any additional logic that can’t
be expressed in the specification itself. Most often, such a class will contain listener methods.
6.2.2 Discarding the component’s body
Components are represented in their container’s templates as a balanced pair of
start and end tags.3 The portion of the template enclosed by those tags is the
component’s body. The body is a mix of static HTML and other components. Figure 6.2 shows how a component’s body is determined. Tapestry divides the page
template into individual chunks, each containing static text, component start
tags, or component end tags. A component’s body is the static text and components it immediately encloses. Enclosed components (such as B in the diagram)
themselves have a body—the static text and components they enclose.
A component may either keep or discard its body. This is controlled by the
allow-body attribute of the <component-specification> element. By default,
allow-body is set to yes, and the component keeps its body (though the corresponding Java class must still include code to render its body). The component
will be able to integrate its body (the portion of the page template its tags
enclose) with its own template (if it has one). This is appropriate for components
Figure 6.2
A component’s body contains
the portion of the template
immediately enclosed by its
start and end tags.
3
So far, most of our examples have been of a page containing a component, but since components may
themselves have templates, they may also be containers of other components. It is not uncommon for
such nesting to go three or four levels deep, and Tapestry has no arbitrary limit on how deep such
component nesting is allowed.
221
such as Conditional or Foreach that want to control if, when, and how often their
body is rendered.
Some components discard their body, usually because the HTML element
they render does not have a body. For example, the Image component discards
its body because the HTML <img> element is always empty.
When Tapestry discards a component’s body, it checks to see if the body contains other components. If so, Tapestry throws an exception (it is not allowed to
define a component only to throw it away).
TIP
If you omit the end tag of a component, the Tapestry template parser
will see an enclosed, discarded component before it sees that the tag is
simply not balanced. You’ll see an exception such as Tag <input> on
line 15 is a dynamic component, and may not appear inside an ignored block.
As you’ll see shortly, a component with its own template can even integrate its
body (which comes from its container’s template) with its own template.
6.2.3 Forbidding informal parameters
In many of our examples from previous chapters, we’ve used informal parameters.
Informal parameters are additional parameters beyond those explicitly declared by
the component (formal parameters are described in the next section).
Most components pass through informal parameters, adding them as attributes
of the HTML element rendered by the component. This is useful in components
such as DirectLink for specifying the CSS class or JavaScript event handlers. The
component doesn’t care about the names or values of the informal parameters; it
passes them through unchanged.
In some cases, a component has no use for informal parameters. For example, the Delegator component doesn’t directly render an HTML element, so it
would have no use for informal parameters. Likewise, the DatePicker component
renders several related HTML elements, so it disallows informal parameters as
well, since it isn’t clear which element would make use of them. Components that
fall into this category will set the allow-informal-parameters attribute to no (the
default for allow-informal-parameters is yes).
When loading a page, the framework checks to see if any informal parameters are bound to a component that does not allow them. An exception is thrown
when informal parameters are used improperly.
To support WYSIWYG editing, there is one exception for components that forbid informal parameters: If an informal parameter is specified in an HTML template as a literal value, the informal parameter is silently ignored.
222
CHAPTER 6
6.2.4 Declaring parameters
The main difference between pages and components is that components may
have parameters and pages may not. Parameters are used to control and customize the behavior of a component to adapt it to the page it is used on. You’ve seen
how common components such as Insert and DirectLink rely on the parameters
you set to make them do the right thing within your pages: render the correct bit
of text (the value parameter of Insert), or invoke the correct listener method (the
listener parameter of DirectLink). This doesn’t apply just to the built-in framework components; the components you create will also require parameters to
adapt them to the specific page and situation they are needed for.
The <parameter> element is used to define formal parameters used by a component. This element includes six attributes, shown in table 6.1.
Table 6.1
The <parameter> element’s attributes
Name
Description
name
The name used for the parameter, which must be a valid Java identifier.
type
The Java type of the parameter, which may be a class name or a primitive type
name.
required
If yes, the parameter must be bound. Defaults to no.
property-name
Optional; specifies the name to use for a connected parameter property when the
default is not allowed.
direction
Specifies how, if at all, you want to connect the parameter binding to a property;
defaults to custom, which means no connected parameter property.
default-value
Optional OGNL expression used as a default value for the parameter if the parameter is not otherwise bound.
Component parameters are more than just JavaBean properties; parameters
may represent a connection, or binding, between a component and some other
object property. A component parameter is not just a property: It is a slot into
which a binding can be plugged. Using the binding, the component can both
read and update the bound property. As an example, consider the value parameter of the Foreach component. Foreach will update the property bound to the
value parameter before rendering its body, for each item in its list (which is provided by its source parameter). Likewise, all the form control elements, such as
TextField, both read and update their value parameters, reading the value
parameter when the page is rendering but updating their value parameter when
the containing form is submitted.
223
The ability to update parameters represents a fundamental difference
between Tapestry component parameters and a typical JavaBean object. A Tapestry component can force a change to another object property using a parameter binding.
As you’ve seen in previous examples, the bindings that are plugged into component parameters run the gamut from short literal strings all the way up to
complex OGNL expressions. At the code level, each parameter is a JavaBeans
property into which a binding object, a class implementing the IBinding interface, is plugged. When a component wants to read or update the value bound to
one of its parameters, the first step is to obtain the matching IBinding object.
Tapestry includes different implementations of IBinding for the three types
of bindings (OGNL expressions, literal static values, and localized messages).4
The IBinding interface includes methods for both reading and updating the
bound value. The getObject() method reads the bound property value. The
setObject() method updates the bound property’s value. Only OGNL expressions can be updated; any attempt to update parameters bound to a static value
or localized message results in a runtime exception.
Figure 6.3 shows how a hypothetical component can access its parameters via
its binding object. The getObject() method retrieves a value through the binding, which can be incorporated into the rendered output of the component.
The binding’s setObject() method is used (on form submission) to update the
bound property.
Before your code can access the bound property value, it needs access to the
binding object itself.
Accessing bindings
For each formal parameter, Tapestry creates a binding property. A binding property is always of type IBinding, and the property name is the same as the parameter name, with Binding appended. As with a property specification, the framework
will use existing accessor methods, but it is easiest to let Tapestry construct a subclass with the accessor methods and the underlying field. To access the binding,
declare and use an abstract accessor:
public abstract IBinding getParameterBinding();
4
A special type of binding is used to access localized messages for pages and components. The details
about this are provided in chapter 7.
224
CHAPTER 6
Figure 6.3
Components use bindings
to get access to properties
of pages or domain
objects. When rendering,
properties are read and
used via getObject().
When rewinding (on form
submission), parameters
are read and converted,
and setObject() is used
to update properties.
Most components do not need to know about bindings, however. Tapestry has
the ability to perform all the interactions with the binding object for you, behind
the scenes, and represent the parameter as a JavaBean property instead. This is
a much easier, more intuitive approach.
Directions for value movements
The direction attribute of the <parameter> element is used to tell Tapestry how
values move between the binding and the component.5 It can be set to one of four
values: in, form, auto, or custom. The majority of component parameters will use
the value in for this attribute. This tells Tapestry to create a connected parameter
5
This is one of the odder naming snafus in the framework. The original intent was based on established
names for parameter directions in languages that allow pass-by-reference: in, out, and in-out. What
the direction attribute has evolved into would better be called “processing” or “connection,” but
names (even bad names) take on a life of their own.
225
property with the same name as the parameter and to manage that property, setting its value at the appropriate time, just before the component renders. As with
specified properties, Tapestry will use existing accessor methods, or create new
accessor methods with this signature:
public Type getParameter();
public void setParameter(Type parameter);
As with specified properties, there is no reason to create your own accessor methods; it is easier to let Tapestry create the accessor methods and underlying fields.
Your code should just declare abstract accessor methods to read, and possibly
update, the parameter property.
If the parameter has a name that conflicts with an existing property (perhaps
one inherited from a base class), you may override the default parameter property name and use another. For example, the PageLink component has a page
parameter, but AbstractComponent already defines a page property, so the
parameter is connected to a different property, targetPage:
<parameter name="page"
type="java.lang.String"
required="yes"
property-name="targetPage"
direction="in"/>
Here, the name of the parameter is still page, but the connected parameter property is named targetPage. The component can access the value bound to the
parameter by defining and using the method getTargetPage(). In addition, a
second property, targetPageBinding, is created to store the binding object for
the parameter (but since the parameter is connected to a property, there’s no
need to access the binding object directly).
The majority of parameters specify in. The processing for such parameters
consists of setting the parameter property from the binding value just before the
component renders, and then resetting the parameter property back to a default
value just after the component renders. Figure 6.4 illustrates the steps taken by
the component to set and clear the properties as part of the component’s render.
In a component, the render() method does three things:
■
It invokes prepareForRender() to prepare the component for rendering.
This sets the value of the direction attribute’s in and form parameters
(reading the value bound to the parameter and updating the connected
property with that value).
■
It invokes renderComponent() to perform the creation of output.
226
CHAPTER 6
■
It invokes cleanupAfterRender() (from within a finally block) to clean up
the component after the render. This reads the direction attribute’s form
parameters and updates the binding value; it also resets the attribute’s in
and form parameters to their default values.
Each component has its own instance of ParameterManager, whose job is to analyze the direction of each parameter and set and reset the connected parameter
properties on the component. This is shown in figure 6.4, where the component
invokes setParameters() on the ParameterManager, which then turns around and
sets the value of connected parameter properties on the component. After connected parameters are set, the component invokes renderComponent() on itself.
Once the component finishes rendering, the properties are reset back to their
initial values by invoking resetParameters() on the ParameterManager—which
again analyzes the parameters for the component and sets the values of the connected parameters back to their initial values.
In some cases, the parameter values, obtained from some domain object, may
represent a large object (or collection of objects) that should be garbage collected. Resetting the parameter value back to null (or whatever the initial value
is) ensures that dangling references to such objects are eliminated and that the
garbage collector can do its work. It also ensures that, for instance, values
entered by one user are not made visible to a different user (in a later request
that uses the same page instance).
The first time ParameterManager’s setParameters() method is invoked, it will
first read each parameter property before setting it from the binding; this value is
saved for later use when the resetParameters() method is invoked. This means
that whatever value is assigned to the property when the component is rendered
the first time will be the reset value used thereafter. A component’s finishLoad()
method can be overridden to set initial values for parameter properties, which will
be used as defaults for the parameters when the parameter is not bound.
Tapestry also tries to optimize the number of properties it sets from parameters. The ParameterManager checks to see if the binding object for each parameter is invariant. Bindings that are either literal values or localized messages are
invariant; that is, unlike bindings based on OGNL expressions, their value will
never change. In these cases, the framework sets the parameter property only
once, just before the component renders the first time. In addition, there is no
need to reset the property after the component renders.
Although most component parameters use in, the other three directions have
their uses:
Figure 6.4 Each component’s ParameterManager is responsible for connected
parameters. For direction="in", the parameter binding is read and the property set
before the component’s renderComponent() method is invoked. After rendering, the
ParameterManager resets the property back to its initial value.
227
228
CHAPTER 6
■
form—The direction attribute’s form value is closely related to in. It
implements the flow appropriate to a form control (such as the value
parameter on a TextField). On render, the property bound to the parameter is read and used to set the connected parameter property value, just
like in. Once the component rewinds, the connected parameter property is
read and the value is used to update the property bound to the parameter,
which in does not do. This direction is used with a component parameter
that is updated from a query parameter upon form submission.
CAUTION
■
Both in and form have a weakness: The parameter property is valid only
while the component is rendering—that is, while the renderComponent()
method is active (or in methods invoked from the renderComponent()
method). There are times when this is insufficient, when the parameter
binding’s value is needed outside the confines of rendering (or rewinding), in which case the value auto is typically used.
custom—The value custom leaves control in your hands. Tapestry creates the
property to store the binding object and nothing else. Your code can access the
binding object and invoke the getObject() and setObject() methods on it.
■
auto—The auto value is much like custom, but it implements a connected
parameter property that appears the same as in. The implementation of
the accessor methods obtains the matching binding object and invokes the
appropriate method upon it, much like this:
public Type getParameter()
{
return (Type)getParameterBinding().getObject("parameter",
Type.class);
}
public void setParameter(Type parameter)
{
getParameterBinding().setObject(parameter);
}
The value auto has some other limitations. The parameter must be required; it
may not be optional. Of the primitive Java types, only boolean, int, and double
are supported.6 (byte, char, float, long, and short are not supported.) In addition, array types are not supported.
6
For types boolean, int, double, and String, the synthetic getter methods use the getBoolean(),
getInt(), getDouble(), and getString() methods of IBinding, respectively. The setter methods
use setBoolean(), setInt(), setDouble(), and setString().
229
Choosing a parameter direction
As the component author, you are responsible for deciding the direction to use
with your parameters, based on how the component and the parameter are utilized. Table 6.2 provides basic rules for choosing the direction.
Table 6.2
Direction
Rules for choosing the direction attribute
Use
in
The parameter is used only when the component is rendering, and the component doesn’t
update the parameter, ever.
form
The component is a form-element component; the parameter is read only when the component is rendering and is updated only when the containing form is submitted. Most of the
form element components include a value parameter with the direction attribute set
to form.
auto
The parameter value may be read or updated even when the component is not rendering (or
rewinding), such as the listener parameter of the DirectLink component.
custom
You’d rather use auto, but you can’t because the parameter type is not valid for auto, or
the parameter is not required.
The vast majority of component parameters are read-only and used only when a
component is actively rendering; in those situations, the value in for the direction attribute is appropriate. Choosing a parameter direction comes down to
identifying when in can’t be used; each direction in table 6.2 should be considered, in order, until a match is found.
It may turn out that a parameter is needed when the component is not rendering. For example, the listener parameter of a DirectLink component is
needed when the DirectLink is triggered; this is an interaction between the
direct service and the DirectLink component that occurs entirely outside the render process (it is shown in chapter 7, figure 7.11).
NOTE
In Tapestry, engine services, such as the direct service, are used to dispatch incoming requests. You can think of them as a kind of servlet inside the Tapestry servlet. Tapestry servlets, engines, and engine services
are described in detail in chapter 7.
The DirectLink component specifies the direction for the listener parameter as
auto, which dynamically accesses the listener via the listener binding. This is a
valid operation even when the DirectLink is not rendering.
230
CHAPTER 6
NOTE
Parameter directions have been identified as an area of Tapestry that
requires work. See appendix A for some notes on how parameters may
be changed and simplified in Tapestry release 3.1.
6.2.5 Reserving names
A potential conflict occurs when a component allows informal parameters: What
happens when the component generates a particular attribute, and the informal
parameter provides a value for the same attribute? For example, the Image component generates an src attribute from its image parameter; but most templates
using an Image component include an src attribute of some kind, to support
WYSIWYG preview. You could end up with the src attribute appearing twice (with
two different values) in the rendered tag, and no clue what the client web browser
will due with this extra attribute.
The solution is to forbid certain names as informal parameters. Each formal
parameter name is automatically added to the list of reserved names that may
not be used as informal parameters. In addition, the <reserved-parameter> element is used to reserve additional names:
<reserved-parameter name="name"/>
Informal parameters that match any reserved name are dropped. The comparison is case insensitive, so the informal parameter Src, SRC, or src would be
dropped from an Image component, which reserves the parameter name src.
When creating components, you should identify which attributes are written
out by your implementation of renderComponent() and reserve each such
attribute as a reserved parameter. All of this applies only to components that
allow informal parameters.
So, now that we have basic tools for specifying components, let’s push on to
see how to combine a specification, a template, and some Java code to create a
fully functional component, using the Hangman application from chapter 2.
6.3 Creating the Digit and Scaffold components
The first of these components will be the Digit component, which is used on the
Guess and Win pages.7 Its input is a numeric digit, and its output, when rendering,
is a complete <img> element that displays the digit as an image. The Scaffold
7
The Home and Lose pages also display a digit, but it is entirely static; so, using ordinary static HTML
is appropriate on those pages.
231
component is similar to the Digit component, the major exception being the choice
of images displayed by the component. Both of these components are codeless,
meaning that they use the BaseComponent class rather than requiring a subclass.
We’ll start by examining the specification for the Digit and Scaffold components, in listings 6.1 and 6.2.
Listing 6.1 Digit.jwc: specification for the Digit component
<!DOCTYPE component-specification PUBLIC
<component-specification
allow-body="no"
allow-informal-parameters="no">
<parameter name="digit" direction="in"
required="yes" type="int"/>
<component id="image" type="Image">
<binding name="image">
getAsset("digit" + digit)
</binding>
<binding name="alt" expression="digit"/>
</component>
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
name="digit0"
name="digit1"
name="digit2"
name="digit3"
name="digit4"
name="digit5"
</component-specification>
Listing 6.2 Scaffold.jwc: specification for the Scaffold component
allow-body="no"
<parameter name="digit" direction="in"
required="yes" type="int"/>
232
CHAPTER 6
getAsset("scaffold" + digit)
</binding>
</component>
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
name="scaffold5"
name="scaffold4"
name="scaffold3"
name="scaffold2"
name="scaffold1"
name="scaffold0"
path="images/scaffold.png"/>
6.3.1 Specifying the digit parameter
The Digit and Scaffold components each take a single parameter, digit, declared
in each component’s specification:
<parameter name="digit" direction="in" required="yes" type="int"/>
The parameter is required; when using either component, a binding for the
digit parameter must be provided or a runtime exception will occur. The type of
the parameter is int, a Java primitive type. This will be the type of the parameter property created to hold the parameter value.
Since the parameter’s direction is in, the framework will automatically set the
parameter property before rendering the component. As you’ll see, we can use this
digit property in the template or as part of OGNL expressions in the specification.
6.3.2 Using the digit parameter
The digit parameter is used only in the specification. The parameter is used in
two ways:
■
To select the correct image asset
■
As the alt attribute for the Image component (providing an accessible title
for the image)
The Image component will be used inside the Digit component’s template. It
contains two parameter bindings:
getAsset("digit" + digit)
</binding>
233
<binding name="alt" expression="digit"/>
</component>
The first binding uses the digit property as part of an OGNL expression to
retrieve the correct asset.8 The digit property is also used with the informal alt
parameter of the Image component. The Scaffold component includes an
almost identical section in its specification:
getAsset("scaffold" + digit)
</binding>
</component>
6.3.3 Creating the template
The Digit component’s template is tiny, just big enough to contain an
Image component:
<img jwcid="image" width="36" height="36"/>
The Scaffold component’s template is similar:
<img jwcid="image" alt="[Scaffold]"
src="images/scaffold.png" border="0"/>
6.3.4 Using the Digit component
Now that we’ve defined the Digit component, we can use it exactly like any of the
components provided with Tapestry. In the revised Guess page’s template, the
Digit component is used instead of the standard Image component:
<img jwcid="@Digit"
digit="ognl:visit.game.incorrectGuessesLeft"
src="images/Chalkboard_3x8.png"/>
As in the previous version from chapter 2, much of the component is simply
there to maintain WYSIWYG preview when using an HTML editor. In this case,
the choice of HTML tag (<img>) and the inclusion of the src attribute are useful
for WYSIWYG preview but are meaningless to Tapestry. Regardless of what
appears in the template, the Image component will render an <img> tag and provide a value for the src attribute. The src informal parameter will be ignored
because it is a literal value provided in the HTML template and the Digit component does not allow informal parameters.
8
This is an example of using the alternate form of the <binding> element, where a long OGNL expression is placed in the body of the element, rather than as the value of the expression attribute. The
form is useful when the expression contains a mixture of quotes, or if it’s very long or complex.
234
CHAPTER 6
This is a marked improvement to what was necessary in chapter 2 to display
the digit. There, the equivalent section of the Guess page template was bogged
down with the details that are now isolated in the Digit component:
<img jwcid="@Image"
height="36"
Not only is the new version more succinct and readable, it also doesn’t require
that the digit assets be specified in both the Win and Guess pages’ specifications.
Those assets are declared just once, inside the Digit component.
6.3.5 Using the Scaffold component
Likewise, the Scaffold component plugs into both the Guess and Win pages in a
similar way:
<img jwcid="@Scaffold"
digit="ognl:visit.game.incorrectGuessesLeft"
src="images/scaffold.png" border="0"/>
Again, the details (and assets) of the component are isolated. In both of these
examples, we’ve seen how the original portion of the Guess page’s HTML template can be extracted and refined as a new component. Our next example takes
a similar approach but does so without using an HTML template of its own.
6.4 Creating the Letter component
The Letter component is used to display an image corresponding to a particular
letter in the alphabet. It is used in several places: on the Guess page, it is used to
display the (partially guessed) target word and to display the remaining guessable letters. The Letter component is also used in the Win and Lose pages to
show the complete target word.
The previous two components (Digit and Scaffold) were codeless and used
HTML templates. The implementation of the Letter component uses no HTML
template; instead, it implements the renderComponent() method to produce the
HTML output in Java code. For well-focused components that produce little
HTML (the Digit, Scaffold, and Letter components each render exactly one tag),
implementing in code is often a better option, especially if there are any calculations to perform as part of the render (in this case, selecting the correct image
based on two different parameters).
Creating the Letter component
235
6.4.1 Specifying the Letter component
The Letter component is similar to the Digit component, taking a letter parameter instead of a digit parameter. It also includes a second parameter, disabled,
which when set to true, forces the component to ignore the letter parameter
and always display a blank space. The specification for this component is provided in listing 6.3.
Listing 6.3 Letter.jwc: specification for the Letter component
class="hangman2.Letter"
allow-body="no"
<parameter name="letter" direction="in"
required="yes" type="char"/>
<parameter name="disabled" direction="in" type="boolean"/>
<context-asset name="space" path="images/letter-spacer.png"/>
<context-asset name="dash" path="images/Chalkboard_5x3.png"/>
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
<context-asset
name="a"
name="b"
name="c"
name="d"
name="e"
name="f"
name="g"
name="h"
name="i"
name="j"
name="k"
name="l"
name="m"
name="n"
name="o"
name="p"
name="q"
name="r"
name="s"
name="t"
name="u"
name="v"
236
CHAPTER 6
<context-asset
<context-asset
<context-asset
<context-asset
name="w"
name="x"
name="y"
name="z"
As with the Digit component, the bulk of the Letter component specification
consists of specifications for all the assets: one for each letter, plus the dash (or
underscore) and a placeholder.
6.4.2 Implementing the Letter component
Much of the code for the Letter component (listing 6.4) is familiar from chapter 2,
where the basic ideas were duplicated in the Guess, Win, and Lose pages. The
code uses the letter parameter to determine the correct asset, and renders an
<img> element, deriving the src attribute from the asset.
Listing 6.4 Letter.java: Java class for the Letter component
package hangman2;
import
import
import
import
org.apache.tapestry.AbstractComponent;
org.apache.tapestry.IAsset;
public abstract class Letter extends AbstractComponent
{
public abstract boolean isDisabled();
public abstract char getLetter();
protected void renderComponent(IMarkupWriter writer,
{
writer.attribute("src",
getLetterImage().buildURL(cycle));
Converts asset
writer.attribute("alt", getLetterLabel());
to URL
writer.attribute("height", 36);
writer.attribute("width", 36);
writer.attribute("border", 0);
}
{
if (isDisabled())
Creating the Letter component
237
char letter = getLetter();
if (letter == '_')
return getAsset("" + letter);
}
{
if (isDisabled())
return " ";
Converts
letter to alt
tag value
return ("" + getLetter()).toUpperCase();
}
}
The main method is renderComponent(), whose job is to render an HTML <img>
tag, which will display the image matching the letter parameter. The IMarkupWriter interface makes it simple to output HTML elements and attributes in an
orderly manner. The beginEmpty() method starts an element that has no body,
and the attribute() methods write attributes for that element, properly quoting
values and converting unsafe characters to HTML entities.
The remaining methods in the Letter class determine the correct image asset
to use based on the letter and disabled parameters, and the proper string to
use as for the <img> tag’s alt attribute. This code was duplicated over several
pages in the original Hangman application in chapter 2, but now it is in a single
place, inside the Letter component.
Previously, you’ve seen how Java code may define abstract accessor methods
to reference specified properties. The same technique is used here to access connected parameter properties. The enhanced subclass generated at runtime by
Tapestry includes the necessary fields, a get method, and a set method. The code
in this class is free to use those accessor methods.
6.4.3 Using the Letter component
The Letter component is used inside the Guess page to generate the grid of
images that may be guessed. As before, this involves a Foreach component (the
selectLoop component) to loop through the available letters, a DirectLink component (the select component) to allow any as-yet-unguessed letter to be
guessed, and an anonymous Letter component to display the actual letter (or a
blank space if the letter has already been guessed):
238
CHAPTER 6
<a href="#" jwcid="select"
class="select-letter"><img jwcid="@Letter"
letter="ognl:letterForGuessIndex"
disabled="ognl:letterGuessed"
border="0"
src="images/Chalkboard_5x3.png"/></a>
</span>
As with the Digit component, using the Letter component is much simpler
than accomplishing the same task using the standard Image component. The
Letter component hides the logic used to select the correct image and label. If
maintaining the ability to preview is not important, then the template could
be simplified further by removing the src and border attributes from the
<img> element.
This is already a simplification from chapter 2, where achieving the same
result required a much more involved section to render the letter image:
<a href="#"
jwcid="select"
class="select-letter">
<img jwcid="@Image"
image="ognl:guessImage"
alt="ognl:guessLabel"
height="36"
width="36"
border="0"/>
</a>
Once again, as with the Digit and the Scaffold components, we’ve consolidated
duplicated logic and duplicated asset specifications, and eliminated duplicated Java code. This makes the revised Guess page even shorter and more
focused on handling guesses rather than figuring out which image to display
for which letter.
Just as we’ve built new components by combining framework components, we
can also build new components from other new components.
6.5 Building the Spell component
Spelling out the target word appears on three out of four pages in the Hangman
application: the Guess page displays the partially guessed word, and both the
Win and Lose pages display the complete target word. In this case, rather than
replace a single Image component with a Letter component, we’ll simplify the
Building the Spell component
239
page template by replacing a Foreach component and an Image component with
a single Spell component.
The earlier versions of these pages used a Foreach component enclosing an
Image component. These components are now the basis for the Spell component, whose HTML template is shown in listing 6.5.
Listing 6.5 Spell.html: HTML template for the Spell component
source="ognl:page.visit.game.letters"
<img jwcid="@Letter" letter="ognl:letter"/>
</span>
This Spell component is hard-wired to do exactly one thing: spell out the word
provided by the Game object’s letters property (via the OGNL expression
page.visit.game.letters). All components have a property, page, which identifies the page that ultimately contains the component.9 From there, we can access
the visit property, as well as the game property of the Visit object.
Because this component is intended for use only within the Hangman application, it doesn’t make sense to define a parameter for the word to spell out; we
will always configure that parameter identically, to display the target word.
Instead, the component is hard-wired to the Game object’s letters property.
6.5.1 Implementing the Spell component
You’ve already seen several examples of page properties being used to coordinate components contained by the page, and this carries over to components as
well, as seen in listing 6.6.
Listing 6.6 Spell.jwc: specification for the Spell component
"-// Apache Software Foundation//Tapestry Specification 3.0//EN"
9
As you’re beginning to see, components can contain components and so forth, so the page may be
several levels of containment above the component.
240
CHAPTER 6
allow-body="no"
<property-specification name="letter" type="char"/>
In this case, the Spell component defines a letter property (of type char), which
allows the Foreach and Letter components contained by Spell to work together.
The Foreach updates the letter property, and the Letter component uses it,
binding it to the letter parameter.
Previously, we’ve discussed how the page acts as controller, mediating between
the components it contains. This concept extends to components (such as Spell)
that contain other components. The container (whether it is a page or just
another component) still is responsible for making it possible for the components it contains to work together. Here, that mediation takes the form of a property that one component (the Foreach) can write into and another component
(the Letter) can read from.
6.5.2 Using the Spell component
The Spell component is used in the Guess page’s HTML template to display the
target word:
<span jwcid="@Spell">
<img height="36" alt="A"
width="36" border="0">
<img height="36" alt="_"
</span>
Earlier, you saw how to add attributes to tags to support previewing. Here, you see
an entire chunk of HTML, the series of <img> elements, used for the same purpose.
Building the Spell component
241
The Spell component does not allow a body, so the entire stretch of HTML within
the Guess page’s template is discarded.
Again, as with the previous components, the new usage is much easier to deal
with than the old one. Back in chapter 2, we had to include the Foreach component, the Image component, and additional methods in our page (again, on all
three major pages: Guess, Win, and Lose). Even omitting the “for preview” <img>
tags (which are ultimately discarded anyway), the old way was much longer:
source="ognl:visit.game.letters"
<img jwcid="@Image"
image="ognl:letterImage"
alt="ognl:letterLabel"
height="36"
width="36"
border="0"/>
</span>
Eventually, a project will hit a crossroads: How valuable is it to maintain the ability to preview the page using a WYSIWYG HTML editor? As a project hits its stride
and pages start to consist of more and more components, the manual effort
needed to maintain the preview (when the HTML is generated within a component) may easily outweigh the utility of doing so. For example, our earlier use of
the Spell component might be replaced with just a placeholder:
<span jwcid="@Spell">
[@Spell component output]
</span>
The page no longer will preview as it did before. Rather than a somewhat accurate snapshot of a running application, the previewed HTML will more obviously
be what it is: a template. Figure 6.5 shows how the updated template will preview
in a WYSIWYG HTML editor.
The result of pulling back the curtain in this way is that HTML developers will
have fewer questions about what they can and cannot change. In addition, HTML
developers won’t be confused as to where changes must take place; there will be
less of a chance that they will make changes to portions of the template that Tapestry is going to discard at runtime. The downside is, of course, that development will gradually shift to the point where the live application must be run in
order to see what final pages will look like.
242
CHAPTER 6
Figure 6.5 Replacing the <img> tags in the Spell component’s body significantly changes how the
template will preview.
6.6 Building the Border component
So far, all the components we’ve built for the Hangman application have discarded their bodies. Exciting possibilities open up when we define components that allow a body, mixing their own template with a portion of their
page’s template.
243
The revised Hangman application uses such a component, Border, to provide
the Shell and Body components for each page in the application. For the Hangman
application, the Border does very little; but in a more realistic Tapestry application,
the same technique can be used to create more of the basic look and feel of the
application: title bars up top, copyrights at the bottom, navigational controls along
the top or side (as appropriate). We’ll see a much more involved example of this in
chapters 9 and 10, with the Border component for the Virtual Library application.
6.6.1 Creating the Border template
The HTML template for the Border component includes a Shell component and
a Body component, and introduces another new component provided by the
framework: RenderBody. The HTML template for the Border component is provided in listing 6.7.
Listing 6.7 Border.html: HTML template for the Border component
<html jwcid="@Shell" title="Tapestry Hangman"
<span jwcid="@RenderBody">
Page specific content goes here.
</span>
</body>
</html>
The RenderBody component causes the component that contains it (the Border
component) to render its body, which is the portion of the page’s template that
its start and end tags wrap around. Note the distinction between being contained
and being enclosed. The RenderBody component is enclosed by the Body component. It is contained by the Border component, because it appears inside the Border component’s HTML template. The Shell, Body, and RenderBody components
are all contained by the Border component.
The body of the Border component may contain any mix of static HTML and
components—even other components with templates (that contain more RenderBody components). RenderBody discards its body, so the note “Page specific
content goes here” will not appear in the live application. Instead, the body of
the Border component within its page is rendered.
244
CHAPTER 6
6.6.2 Creating the Border specification
The Border component’s specification is shown in listing 6.8.
Listing 6.8 Border.jwc: specification for the Border component
allow-body="yes"
<context-asset name="stylesheet" path="css/hangman.css"/>
This specification serves two simple purposes:
■
It indicates that the Border component allows, and does not discard,
its body.10
■
It allows us to specify here, and only here, the asset for the application’s stylesheet.
Once again, using a component has allowed us to remove duplicate code, duplicate specifications, and duplicate assets. As usual in Tapestry, simplicity and consistency go hand in hand: Using the Border component is simpler, more
consistent, and more maintainable than duplicating the Shell and Body components and the stylesheet asset.
6.6.3 Using the Border component
Using the Border component is the same as using any other component. You
create a <span> tag in the HTML template to indicate where on the page the
component belongs and to delineate the body of the component. All four pages
in the application follow the same general template:
<html>
<head>
10
This could be omitted, since the default is yes, but including the attribute as a reminder helps to document the purpose of the component.
245
</head>
<body jwcid="$content$">
<span jwcid="@Border">
. . .
</span>
</body>
</html>
Back in chapter 2, we introduced the special jwcid "$remove$", used to edit out a
portion of the template. When using the Border component, we have a similar
problem. At runtime, we want the Border component (and the Shell and Body
components it contains) to render the <html>, <head>, and <body> elements in
the page. This creates a conflict if we want to maintain the ability to preview the
page in the HTML editor; we need to keep the <link> tag that connects the template to the stylesheet, or the page becomes virtually unreadable in the editor.
At the same time, we want to eliminate the <html>, <head>, and <body> elements in the page template, because they will be provided at runtime by the Border component (via the Shell component contained by the Border). If we use the
"$remove$" jwcid on the <html> or <body> tags in the page template, the entire
page will be removed! The solution is a second special jwcid, "$content$".
The "$remove$" jwcid is used to remove an element and its body from the
template. The "$content$" jwcid removes everything but the element’s body.
The <body> tag, everything that precedes it, the </body> tag, and everything that
follows it are all discarded. What’s left is the content inside the <body> tag, which
is the Border component, enclosing everything else on the page. Everything
then falls into place, and at runtime, the Border component properly renders
the <html> and <body> tags before and around its body.
The end result of all these changes is shorter, simpler Java code and smaller,
simpler templates. The code base for the entire application shrank from approximately 200 lines of code (excluding comments) to 160 lines, but, more important,
virtually all redundancy in the code was removed. As much as application-specific
components such as these are a modest win for an individual application, the real
gains are to be made when creating fully reusable components, adaptable to many
applications. The next sections describe those sorts of components.
246
CHAPTER 6
6.7 Creating interactive, reusable components
The components described in the previous sections are output-only components:
They are used when a page is rendering, but they are not themselves interactive.
Adding interactivity requires using components in combinations with engine services. Engine services, described more fully in chapter 7, are the gateways connecting incoming requests from the client web browser with objects and code running
on the server.
As an example, let’s implement a clickable link for an image map. The
example in this section is part of the examples application. Launch your web
browser to http://localhost:8080/examples/app and click the Pets link under the
chapter 6 heading.
6.7.1 Introducing the Pet Store image map
Image maps are an often-overlooked aspect of HTML. An image map is an ordinary image paired with additional HTML to define regions within the image.
Each region is a rectangle or polygon within the image, defined using a series of
coordinates. Each region defines a URL to trigger when that region of the
image is clicked. HTML designers often use image maps because of the precision with which they can arrange buttons (or other clickable elements) inside
the image.
To put this all together, the <img> element includes a usemap attribute that
identifies the image map to use. A corresponding <map> element contains a number of <area> elements, and each <area> defines a region of the image and the
URL to trigger.
A familiar example of this is part of the standard Java 2 Enterprise Edition
(J2EE) demonstration application, the Java Pet Store. The first page of the Pet
Store includes an image map for selecting the top-level category when shopping
for a pet. Clicking the pet image brings up a catalog of that type of pet. A Tapestry version of this page is shown in figure 6.6. The utility of an image map is
obvious here because the clickable regions (the small images of the different
pets) are not organized into handy rows and columns.
It would be ideal to leverage the functionality of the DirectLink component
(the component used so effectively in the Hangman application) but use it
instead to create the <area> elements that define the clickable regions. We must
still be able to define listener methods and pass service parameters as part of the
URL. This ideal is possible, because most of the behavior of the DirectLink component is inside the direct service, and so it is easy to reuse in a new component.
Creating interactive, reusable components
247
Figure 6.6 An image map allows the user to select a category of pet to shop
for. Image maps allow for precise control over the layout of the active areas.
The DirectArea component can use the direct service just as easily as the
DirectLink component.
6.7.2 Specifying the DirectArea component
The DirectArea component, whose specification is shown in listing 6.9, is similar
to a DirectLink component; it has a required listener parameter and an
optional parameters parameter. Like the DirectLink component, the DirectArea
component must support informal parameters.
Listing 6.9 DirectArea.jwc: specification for the DirectArea component
248
CHAPTER 6
class="examples.DirectArea"
allow-body="no"
allow-informal-parameters="yes">
<parameter name="listener" required="yes"
direction="auto"
type="org.apache.tapestry.IActionListener"/>
<parameter name="parameters" type="java.lang.Object"
direction="in"/>
b
c
b
c
Listener methods are represented as instance of the class IActionListener.
We use java.lang.Object as the type since we don’t know the actual type that will
be provided.
The HTML <area> element does not allow a body, so neither does the DirectArea
component. The specification defines the two parameters, listener and parameters, mimicking two of the DirectLink parameters. The listener parameter’s
direction is auto, not in. This is necessary because the listener is needed when
the direct service triggers the component, and this event occurs when the component is not rendering.
In addition, the parameters parameter is type java.lang.Object; this is the
correct type to use when the runtime type of the parameter is not known. As with
the DirectLink component (which defines its parameters parameter the same
way), this value may be an object, an array of objects, or a List of objects.
6.7.3 Implementing the DirectArea component
The DirectArea component’s class, shown in listing 6.10, consists of only three
methods. The renderComponent() method renders the <area> element. The
trigger() and isStateful() methods are used when a request is sent by the client web browser in response to the user clicking within the image.
Listing 6.10 DirectArea.java: Java class for the DirectArea component
package examples;
import org.apache.tapestry.AbstractComponent;
import org.apache.tapestry.IActionListener;
import org.apache.tapestry.IDirect;
import
import
import
import
import
import
249
org.apache.tapestry.Tapestry;
org.apache.tapestry.engine.IEngineService;
org.apache.tapestry.engine.ILink;
org.apache.tapestry.link.DirectLink;
public abstract class DirectArea
extends AbstractComponent
implements IDirect
{
public abstract IActionListener getListener();
public abstract Object getParameters();
b
{
if (cycle.isRewinding())
return;
Object[] parameters =
DirectLink.constructServiceParameters(getParameters());
IEngineService service =
cycle.getEngine().getService(Tapestry.DIRECT_SERVICE);
ILink link = service.getLink(cycle, this, parameters);
writer.beginEmpty("area");
writer.attribute("href", link.getURL());
renderInformalParameters(writer, cycle);
c
d
e
f
}
public void trigger(IRequestCycle cycle)
{
IActionListener listener = getListener();
g
if (listener == null)
throw Tapestry.createRequiredParameterException(this,
"listener");
listener.actionTriggered(this, cycle);
i
}
public boolean isStateful()
{
return false;
}
j
h
250
CHAPTER 6
b
c
d
e
f
g
h
i
j
The direct service, part of the Tapestry framework, requires that the component
it invokes implement the IDirect interface.
The DirectLink class provides a static utility method, constructServiceParameters(), for converting the value bound to the parameters parameter into
an array of objects.
The IEngineService instance is responsible for constructing the URL, which is
provided as an instance of ILink.
The link object can provide the URL for this component.
The AbstractComponent base class provides a method, renderInformalParameters(),
for generating additional HTML attributes from informal parameters bound to
the component.
The trigger() method is invoked by the direct service (it is part of the
IDirect interface).
This code gets the listener object (an object implementing the IActionListener
interface that will invoke the listener method) and checks that the object exists
and is not null.
The actionTriggered() method is defined by the IActionListener interface; the
listener object will invoke the listener method.
The isStateful() method is also defined by the IDirect interface and invoked
by the direct service.
The core of this class is the renderComponent() method. It starts with a simple
optimization; if the request cycle is rewinding (to process a form submission)
instead of rendering, then there’s no point in doing the rest of the work (any output will be discarded during a rewind). This scenario could occur if a Form happens to enclose the DirectArea component. When the Form is submitted, the
rewind phase will pass through the DirectArea as well as all the form control
components, but the DirectArea component has nothing to contribute to the
Form’s rewind.
The next step is to convert the parameters parameter into an array of objects.
A static utility method in the DirectLink class accomplishes this for us. It will
analyze the object and return an array of objects (if the value is an array or a
List). The method will return null if there are no service parameters.
Next, we obtain the direct service from the engine and have the service construct a link back to this particular DirectArea component, including any service
parameters. The next few calls to the markup writer create the <area> element
and set the href attribute from the link created by the service. This component
must support informal parameters; the call to renderInformalParameters() will
251
convert any informal parameters into additional attributes. Components that
allow a body might invoke the method renderBody() as well, but DirectArea
doesn’t allow a body (because <area> elements are always empty).
The flip side of the renderComponent() method is a pair of methods, defined
by the IDirect interface, that are invoked when the link (encoded into the <area>
element’s href attribute) is triggered. As we’ll discuss in chapter 7, the direct service will locate this component and start by invoking the isStateful() method.
The isStateful() method in DirectArea always returns false, indicating that
the link is not stateful, and that it is okay to execute cold (when the HttpSession
doesn’t exist or has expired). That’s acceptable for this example, but a truly reusable DirectArea component would, like the DirectLink component, provide a
parameter for controlling the return value for this method.
The direct service then invokes the trigger() method. Invoking getListener()
will retrieve the listener object from the binding for the listener parameter.
Despite the fact that this parameter is required, it’s still possible that the listener
is null. The required attribute on the <parameter> element only ensures that the
parameter was bound, not that the bound property is non-null. For example,
using the value ognl:null would satisfy the parameter’s required check, but it
wouldn’t give us a useful value. To be safe, we check that the listener object is not
null and throw an exception in the unlikely case that it is. The Tapestry class (a
utility class containing useful static methods) includes a method for constructing
this exception.
You can also see why it was necessary to set the listener parameter’s direction attribute to auto instead of in. This trigger() method is invoked when the
DirectArea component is not rendering—the semantics of an in parameter
would not work, since a connected property for an in parameter will have the
correct value only while the renderComponent() method is executing. The direct
service invokes the trigger() method in an entirely different request, at a different time, when the renderComponent() method is definitely not executing.
Using auto for the listener parameter’s direction attribute is appropriate.
Invoking the getListener() method will get the listener object from the binding,
without depending on the renderComponent() method.
After successfully passing the required check, we know we have a non-null listener and can invoke the actionTriggered() method on the listener object. This
will ultimately invoke the listener method provided by the application.
252
CHAPTER 6
6.7.4 Using the DirectArea component
The Pets page uses several instances of the DirectArea component. The majority of
the template, shown in listing 6.11, is simple, static HTML for the image and for
the <map> element referenced by the image. Only the <area> elements are
dynamic; each of these is a DirectArea component with its own listener.
Listing 6.11 Pets.html: HTML template for the Pets page
<span jwcid="@Border" title="Pets">
<p>Select a type of pet to continue:</p>
<img src="images/pets-image-map.gif"
alt="Pet Selection Map"
usemap="#petmap"
width="350"
height="355"
border="0"/>
<map name="petmap">
<area jwcid="@DirectArea" listener="ognl:listeners.selectBirds"
alt="Birds"
coords="72,2,280,250"/>
<area jwcid="@DirectArea" listener="ognl:listeners.selectFish"
alt="Fish"
coords="2,180,72,250"/>
<area jwcid="@DirectArea" listener="ognl:listeners.selectDogs"
alt="Dogs"
coords="60,250,130,320"/>
<area jwcid="@DirectArea" listener="ognl:listeners.selectReptiles"
alt="Reptiles"
coords="140,270,210,340"/>
<area jwcid="@DirectArea" listener="ognl:listeners.selectCats"
alt="Cats"
coords="225,240,295,310"/>
<area jwcid="@DirectArea" listener="ognl:listeners.selectBirds"
alt="Birds"
coords="280,180,350,250"/>
</map>
</span>
Inside the Java class, each listener method funnels into a single method that
configures the PetCategory page and selects it as the response page:
private void select(String type)
{
IRequestCycle cycle = getRequestCycle();
253
PetCategory next = (PetCategory) cycle.getPage("PetCategory");
next.setType(type);
cycle.setPage(next);
}
public void selectBirds(IRequestCycle cycle)
{
select("birds");
}
This brings us up to the PetCategory page, which now has the information it
needs to display the page, customized to the user’s selection. An example is
In a real application, there would be a product database, and the PetCategory
page would perform a query and produce a listing of dogs in the inventory. For
the purposes of this example, just getting to the page (and acknowledging that
the user selected dogs) is sufficient.
Creating components with interactivity is nearly as effortless in Tapestry as
creating simple output-only components. Because of the way the framework
manages both the component object graph and the dispatch of incoming
requests, it is easy to create links that invoke methods on component objects.
From there, it is simple to call back into application-specific code using listener methods.
Figure 6.7 After selecting a pet using the image map, the user is brought to a page that displays
an inventory of pets of the selected type.
254
CHAPTER 6
Creating components specific to an application is an excellent way to leverage
the various capabilities of members of a team. Because components are, by their
nature, small and self-contained, they represent tasks that are more accessible to
less-experienced team members, or to team members with specialized knowledge (of Tapestry, HTML, or client-side JavaScript).
Although we’ve discussed creating components that are reusable in multiple
applications, we haven’t covered the nuts and bolts of actually doing so. You
could cut and paste the component’s class, specification, and template into a new
project, but then you’d be faced with all the problems of maintaining the same
source code in multiple locations. Instead, Tapestry has a sophisticated way of
packaging components together into libraries, as you’ll see in the next section.
6.8 Using component libraries
So far, you’ve seen how to use components packaged with the Tapestry framework, and you’ve seen several techniques for creating new, application-specific
components. The framework and the application are not the only places that
components can come from: Components can be packaged into libraries and
distributed as Java Archive (JAR) files.
The Tapestry distribution includes a secondary library, tapestry-contrib-3.0.jar.
This contains the Contrib (short for Contributions) library, which includes many
useful components that are not absolutely required by the framework itself. In
addition, this library contains components that have been contributed by outside
developers. Over time, components can migrate from the Contrib library into
the main framework.
There are no special requirements for using component libraries (beyond
declaring them, as described in the next section). The JAR containing the library
must be on the classpath; this can be accomplished easily by copying the JAR into
the WEB-INF/lib folder of the web application.
6.8.1 Declaring libraries
Before you may use a library in an application, you must first declare its use for
your application. This requires that your application have an application specification. Application specifications are optional; applications that don’t use special features (such as component libraries) don’t need an application specification.
The application specification has the file extension .application and is stored
in the WEB-INF folder, along with the page and component specifications. The
name of the file matches the name of the servlet. For most of the examples in
Using component libraries
255
this book, the name of the servlet is examples. This is shown in the web application deployment descriptor (see listing 6.12).
Listing 6.12 web.xml: web deployment descriptor for the example application
<!DOCTYPE web-app PUBLIC
"-//Sun Microsystems, Inc.//DTD Web Application 2.2//EN"
"http://java.sun.com/j2ee/dtds/web-app_2_2.dtd">
<web-app>
<servlet>
<servlet-name>examples</servlet-name>
<servlet-class>org.apache.tapestry.ApplicationServlet
</servlet-class>
<init-param>
<param-name>org.apache.tapestry.visit-class</param-name>
<param-value>examples.Visit</param-value>
</init-param>
</servlet>
<servlet-mapping>
<servlet-name>examples</servlet-name>
</servlet-mapping>
Provides mapping to
servlet from URL
</web-app>
Because the name of the servlet is examples, Tapestry will expect the application
specification to be the file WEB-INF/examples.application. This file is shown in
listing 6.13.
Listing 6.13 examples.application: application specification for the examples application
<!DOCTYPE application PUBLIC
<application name="Tapestry in Action Examples">
<library id="contrib" specification-path=
"/org/apache/tapestry/contrib/Contrib.library"/>
</application>
256
CHAPTER 6
The root element for an application specification is <application>. You can
optionally provide a user-presentable name for the application, which is used in
some error messages. The <library> element defines an ID for a library and
provides the path to the library’s specification. The library ID must be a valid
Java identifier. An application may use any number of libraries, each with a
unique ID.
Application specifications can also be used to override the default locations
for page and component specifications, but there’s rarely a need to do so. The
reference in appendix D has the details.
The specification path for a library is the complete path to the library’s specification. This is a path within the Java classpath, which is appropriate because
libraries are packaged as JAR files, with their resources available only through
the classpath.
Now that we’ve given this library an ID, we can start using the components
within the library.
6.8.2 Referencing library components
Components in libraries are referenced using the library ID as a prefix—for
example, contrib:Table or contrib:Palette (for components Table and Palette
within the component library with the ID contrib). This extended naming pattern will be familiar to anyone who has used XML namespaces. Prefixing is
important because different libraries may define different components with the
same type name. The prefix allows both you and Tapestry to tell one component
from another.
Using a component from a library is as easy as using a component provided
with the framework, or one particular to your application. Figure 6.8 shows an
example of this. This page includes the Palette component, a sophisticated component available in the Contrib library. A Palette component is used for handling multiple selections of items from a list and (optionally) allows the user to
order the selections in the list.
The Palette component represents the upper potential of fully reusable
components. It combines sophisticated client-side JavaScript with images (distributed with the component inside the JAR file), and yet the look and feel of the
Palette may be customized by changing the page’s stylesheet or by overriding
the default images used for the buttons (in the center column). Despite all this, the
component is no more difficult to use than the PropertySelection component
described in chapter 4.
257
Figure 6.8 The Palette component allows multiple selection and ordering. The buttons in the center
are used to move items between the two lists and to reorder items in the selected list (the right
column).
Creating the Resume Builder
The example page for the Palette component is the Resume Builder, a page that
could be drawn out of an online job search program. In the template (see
listing 6.14), the Palette component is represented (arbitrarily) as a <select> element.
Listing 6.14 ResumeBuilder.html: HTML template for the Resume Builder page
<body jwcid="@Border" title="Resume Builder">
b
<span class="title">Resume Builder</span>
<p>Please fill out a few questions so that
we can compile your on-line resume.
</p>
258
CHAPTER 6
<span jwcid="@FormError" delegate="ognl:beans.delegate"/>
<form jwcid="@Form" listener="ognl:listeners.formSubmit"
<tr>
<th>Experience:</th>
<td><select jwcid="inputExperience"/> Years</td>
</tr>
c
<tr class="tall">
<th><span jwcid="labelLanguages">Languages</span>:</th>
<td><select jwcid="inputLanguages"/>
d
<p>
Select the programming languages you are proficient in,
and order them by relative level of skill.
</p>
</td>
</tr>
<tr>
<td>
<input type="submit" value="Continue"/>
</td>
</tr>
</table>
</form>
</body>
b
c
d
Several of the pages in the examples application make use of a Border component to deal with stylesheet issues.
This <select> is a PropertySelection component.
The Palette component is also represented by a <select> tag.
For comparison, the Resume Builder page includes a PropertySelection component as well as the Palette. Both are represented in the template as <select> tags,
even though the Palette component will generate, at runtime, a considerable
amount of HTML , including a <table> (to control layout), several links and
images, and two different <select> elements. The downside to using sophisticated
259
components such as Palette is that doing so unavoidably undermines the WYSIWYG
preview of the page template.
Declaring components in the page specification
Both the PropertySelection and Palette components are declared components,
with their type and parameters provided in the page specification. Listing 6.15
shows the specification for the Resume Builder page, which contains <component>
elements for these components.
Listing 6.15 ResumeBuilder.page: specification for the Resume Builder page
<page-specification class="examples.resume.ResumeBuilder">
<bean name="delegate"
class="examples.register.RegisterDelegate"/>
<property-specification name="experience"
type="examples.resume.Experience">
@examples.resume.Experience@JUNIOR
</property-specification>
b
<property-specification name="languages"
type="java.util.List"/>
<component id="inputExperience" type="PropertySelection">
<binding name="model" expression="experienceModel"/>
<binding name="value" expression="experience"/>
</component>
<component id="labelLanguages" type="FieldLabel">
<binding name="field" expression="components.inputLanguages"/>
<static-binding name="displayName" value="Languages"/>
</component>
<component id="inputLanguages" type="contrib:Palette">
<binding name="selected" expression="languages"/>
<binding name="model" expression="languageModel"/>
<binding name="sort">
@org.apache.tapestry.contrib.palette.SortMode@USER
</binding>
</component>
d
c
e
f
260
CHAPTER 6
b
c
d
e
f
A default value can be specified for properties (as an OGNL expression). This
expression references the public static field JUNIOR of the Java class examples.
resume.Experience.
The contrib: prefix references the library declared in the application specification.
The selected parameter is a List of selected options.
Palette uses an instance of IPropertySelectionModel to define the possible
options, just as the PropertySelection component does.
SortMode.USER allows for manual ordering of the selected items.
Finally, we see that library ID prefix as promised; the component type for the
inputLanguages component is contrib:Palette, the Palette component provided in the contrib library. This same syntax can be used with implicit components; for example, the inputLanguages component could be declared entirely
in the HTML template as
<select jwcid="inputLanguages@contrib:Palette"
selected="ognl:languages"
model="ognl:languageModel"
sort="ognl:@org.apache.tapestry.contrib.palette.SortMode@USER"/>
A Palette component is configured in much the same way as a PropertySelection
component. The possible values for a Palette component are defined using the
same IPropertySelectionModel interface that is used by the PropertySelection
component. Unlike PropertySelection, the Palette component is used to edit a
java.util.List of values, bound to its selected parameter. In this example, we
create a languages property on the page and have the Palette edit that property.
Part of the configuration for a Palette is the order in which items are displayed in the two columns. Items may be sorted by the label (as displayed in the
user interface) or by the hidden value of each item. In this example, a third
option is used, which allows manual sorting of the selected values. All of this is
configured through the Palette’s sort parameter.
Using the page stylesheet
Part of the Palette’s user interface relies on the stylesheet of the page. The Palette component renders an HTML <table> element to lay out the two columns,
the headers, and the column of buttons in the middle. This <table> is (by
default) given the CSS class tapestry-palette . The stylesheet for the page
includes several entries for properly displaying the Palette:
TABLE.tapestry-palette TH
{
background-color: black;
261
color: white;
font-weight: bold;
text-align: center;
}
TABLE.tapestry-palette SELECT
{
font-weight: bold;
width: 200px;
}
TABLE.tapestry-palette TD.controls
{
text-align: center;
vertical-align: middle;
width: 60px;
}
The HTML generated by the Palette merges with these entries in the stylesheet
to produce the desired look and feel. The first block sets the headers (Available
and Selected) as white text on a black background. The second block fixes the
width of the <select> element lists. The last block sets the width of the column
containing the buttons (forcing the buttons to stack vertically) and centers them
horizontally and vertically. This style of CSS “hook” allows for considerable flexibility in customizing the look and feel of the Palette to fit seamlessly into the
page that contains it. Without this support from the page’s stylesheet, the Palette
would still operate but would not look as pleasing; this is demonstrated in
figure 6.9.
The remainder of the implementation of the Resume Builder page is standard form processing, as covered in chapter 4. What’s important is not all the
features of the Palette component, but how easily such a complex component can
be incorporated into your application.
So, using components from a library is simple enough—and, as you’ll see next,
packaging your own components into libraries is nearly as easy.
6.9 Packaging components into libraries
The next step beyond creating a reusable component is to package that component inside a library for later reuse. Creating a component library is a very
straightforward process; it just involves rethinking where some of the resources
for the component (the specification, the template, and any assets) are stored. In
a component library, everything is stored in the classpath, whereas application
resources are stored within the web application context.
262
CHAPTER 6
Figure 6.9 Without the support of the page’s stylesheet, the Palette is functional but not as
visually pleasing.
NOTE
Appendix B describes the layout and process for building a Tapestry application using the Ant build tool. Building a component library as a JAR
is structurally similar to building a Tapestry web application as a Web Archive (WAR) file. In the examples distribution, the examples-library
subproject shows how to build such a library (containing the components
described here and in chapter 8). The banner ads and examples subprojects show how to include the library as part of a web application.
6.9.1 Creating the library specification
Constructing new components to support your application is as easy as creating
pages; in fact, it is largely the same process. Compartmentalizing your application
in this way helps to remove redundancy in your code and specifications. It’s also a
good way of dividing up your team, with “big picture” developers working on
263
pages, critical components, and back-end systems, and less-seasoned developers
constructing simple components for the page developers. Tapestry’s approach
allows the application to be divided into small pieces that can be assigned to developers with just the right skill level.
Once you’ve created your own components or acquired new components in a
component library, using them in your own application is as simple and natural
as any of the components provided with the Tapestry framework.
To illustrate this process, we will revisit the Resume Builder example from the
previous section. The Resume Builder page includes form-level validation to
ensure that the user supplies at least one language proficiency. Figure 6.10 shows
the page with a warning to that effect.
Figure 6.10 The Resume Builder page uses a FormError component (from a component
library) to display the form-level validation.
264
CHAPTER 6
In the previous section, the FormError component was an application-specific
component. Let’s convert it into a library component instead.
The first step is to select a Java package for the library and the components
within the library. For this example, the Java package is examples.library. In this
package directory, we will store the library specification, the component specification, and the component template, as well as a related asset (the X icon used
by the component).
Like an application specification, the library specification can declare additional libraries used by this library. The library specification for this example
library doesn’t use this facility, so the specification is simply a placeholder:
<!DOCTYPE library-specification PUBLIC
<library-specification/>
The library specification is named Examples.library and is in the examples.
library package. In the examples distribution, it is stored in the directory exampleslibrary/src/java/examples/library. In chapter 8, we’ll discuss some of the additional elements that may be put into a library specification; for now, this specification is a placeholder. By giving the library a particular location on the classpath
(/examples/library), Tapestry can find any components that are included in the
library, because they’ll be in the same location.
6.9.3 Creating the FormError component
The FormError component is created in the same directory as the library specification. In fact, any component specifications in the directory with the library
specification will be components that are part of the library.
The FormError component will take, as a parameter, an IValidationDelegate
instance. The HTML template for the component is shown in listing 6.16.
Listing 6.16 FormError.html: HTML template for the FormError component
<span jwcid="@Conditional" condition="ognl:delegate.hasErrors">
<table style="{
border-style : solid;
color : Red;
265
font-weight : bold;
background-color : #FFFF40;
border-width : 1px;
border-color : Black;
padding : 2px;
}">
<tr valign="top">
<td>
<img jwcid="@Image" height="52" alt="[Error]"
image="ognl:assets.icon" width="52"/>
</td>
<td>
delegate="ognl:delegate.firstError">Error Message</span>
</td>
</tr>
</table>
</span>
In order to package the FormError component as a template, some minor
changes had to be made from the original page. In the original template, the
outer <table> element referenced the CSS style for the table indirectly:
The stylesheet for the page included the following in the declaration:
TABLE.error {
border-style : solid;
color : Red;
font-weight : bold;
background-color : #FFFF40;
border-width : 1px;
border-color : Black;
padding : 2px;
}
For this component to be fully reusable, it may not rely on any particular information in the page’s stylesheet.11 The page may not have a stylesheet, or it may
have a stylesheet that doesn’t include the TABLE.error entry. Instead of referencing the stylesheet, the FormError component includes the necessary style information inline.
11
Another option is to document the stylesheet requirements made by the component and force the developer to include the necessary entries in the page’s stylesheet. The Palette component partially
adopts this approach.
266
CHAPTER 6
Referencing packaged images
The icon image (the X symbol) poses a similar problem: The library may be distributed as a JAR and should work with any application it is dropped into, but it
also is reliant on the image asset, form-error.png. This file is packaged with the
component inside the component library JAR; it is a resource on the classpath,
not normally visible to the client web browser.
Normally, the client web browser may only access files (static HTML files,
stylesheets, images, and so forth) that are stored in the context. Anything stored
in a JAR file, or under the WEB-INF folder, is not accessible. That begs the question: How is the client web browser ever going to gain access to the file? Certainly, we don’t want any solution where the files are packaged separately or must
be unpackaged from the JAR files to be useful; that goes against the plug-andplay nature of Tapestry.
Tapestry includes a mechanism for serving images and other assets directly
from the JAR file. You’ll see how to declare such assets shortly, but the difference
in the HTML template is the need to use an Image component to display the X
icon rather than simple static HTML.
The component specification for the FormError defines the delegate parameter
and the asset for the error icon. The specification is named FormError.jwc (see
listing 6.17), and it is stored (with the library specification) in the /examples/
library folder.
Listing 6.17 FormError.jwc: specification for the FormError component
allow-body="no"
<parameter name="delegate" direction="in" required="yes"
type="org.apache.tapestry.valid.IValidationDelegate"/>
<private-asset name="icon" resource-path="form-error.png"/>
267
A new specification element, <private-asset>, is used (instead of <contextasset>) to declare assets that are stored on the classpath rather than in the web
application context. Private assets are localized, just the same as ordinary context
assets. Components (such as the Image component in the FormError component’s template) do not know or care whether assets are context or private.
Private assets work with the asset service (one of the eight default services
provided by the framework) to expose these files to the client web browser. At
runtime, rather than build a URL to a static file within the web application context, the asset service builds an application URL. When the URL is triggered, the
asset service will locate the asset on the classpath and send its contents, as a byte
stream, back to the client web browser.
If you examine the source for the page when an error message is displayed,
the Image component renders as
<img src="/examples/app?service=asset&sp=S%2Fexamples%2Flibrary
%2Fform-error.png"
border="0" width="52" alt="[Error]" height="52"/>
This URL invokes the asset service to retrieve the contents of a file stored somewhere on the classpath. In this example, the location of the file within the
classpath is /examples/library/form-error.png (the %2F is an encoding of the
slash character).
TIP
Tapestry can also be configured to dynamically copy the contents of private assets to a web folder directory and build a static URL pointing to
the resource. This is a better approach in production, since web servers
are much faster at servicing requests for static resources than handling
dynamic requests involving servlets.
6.9.4 Using the FormError component
Using the FormError component is the same as using any other component. The
library containing the component must be declared in the application specification:
<library id="lib"
specification-path="/examples/library/Examples.library"/>
This declaration identifies a second library used by the application that can be
referenced with the prefix lib:. In the ResumeBuilder template, the FormError
component is accessed using that prefix:
<span jwcid="@lib:FormError" delegate="ognl:beans.delegate"/>
268
CHAPTER 6
Because of the prefixing system used, a single page can combine components
provided by any number of libraries. The Resume Builder page, for example,
still uses the contrib:Palette component later in its template.
6.10 Summary
Building reusable components in Tapestry is an extension of the techniques
you’ve already seen for building pages. Because creating components is so easy in
Tapestry, it becomes a natural problem-solving technique: divide (into components) and conquer. Building components is a great way to organize your application; it removes redundancies from your templates, specifications, and Java code.
Tapestry includes a simple and elegant way to create completely reusable components and package them as libraries. The components you create today for your
current application can be instantly reused tomorrow, in your next application.
The framework even allows for libraries to build on each other; new components can be assembled by combining components provided by the framework
or by other libraries.
Creating new components is just one approach to extending the basic capabilities of the Tapestry framework. In chapter 7, we’ll go into more detail about
the internals of the framework and then utilize that information to build even
more powerful components in chapter 8.
This chapter covers
■
Managing server-side state
■
Understanding clustering
■
Creating URLs and dispatching requests
■
Rendering pages in detail
■
■
Localizing applications
269
270
CHAPTER 7
Many applications can be built in Tapestry using just the techniques described in
the previous chapters and an understanding of the existing components. In
preparation for even more ambitious things (in chapter 8), this chapter covers a
bit more about how Tapestry operates internally, describing what happens
between the time a request is received by the application server and the time a
listener method is invoked. We also touch on issues related to server-side state
management, application localization, and how Tapestry makes use of object
pools for efficiency. All of this background material lays the groundwork for
understanding how to create more involved applications, as well as how to create
more powerful, reusable component libraries and how to extend and change the
basic behavior of the Tapestry framework.
7.1 Processing requests
In earlier chapters, we talked about offloading complexity into the framework by
allowing it to control the request/response processing cycle. In this chapter, we
describe how Tapestry processes the incoming request. Processing requests
involves a number of overlapping concepts:
■
Parsing and caching page specifications, page templates, component specifications, and component templates
■
Pooling instances of pages
■
Managing server-side state, including persistent page properties
■
Localizing text and images on pages
■
Interpreting incoming requests and dispatching to the correct objects and
methods
Table 7.1 lists the key objects responsible for processing requests.
Table 7.1
Request processing framework objects used to dispatch incoming requests
Object
Description
Application servlet
Serves as a bridge between the Servlet API and Tapestry
Engine
Acts as a central hub for request processing, managing of server-side state, and
accessing of Tapestry subsystems
Engine service
Builds and interprets URLs for specific Tapestry operations
Request context
Creates a façade over Servlet API objects (HttpServletRequest,
HttpSession, etc.)
Table 7.1
271
Request processing framework objects used to dispatch incoming requests (continued)
Object
Description
Request cycle
Coordinates processing of a single-request, temporary cache for page instances
Page recorder
Manages persistent page properties for a single page; moves data between page
properties and HttpSession
Page source
Manages an object pool used by the request cycle to obtain page instances
Page loader
Creates and configures new page instances as needed
7.2 Understanding the application servlet
In Tapestry, every application uses exactly one servlet: the application servlet. In
traditional servlet applications there are many servlets, each representing a different operation that may occur in the application. Each of these servlets is new
code for you to write, test, and debug.
One of the reasons writing servlets is more involved than writing most other
kinds of Java applications is that servlets operate from within a difficult environment. Exactly one instance of each servlet is created by the servlet container.
That single servlet may handle dozens of simultaneous requests, each in a separate thread. This multithreading makes instance variables on the servlet almost
useless. There’s no point in storing any information about the current client in
an instance variable; some other thread is sure to overwrite your change almost
immediately. All too frequently, subtle bugs creep in where developers violate
these basic rules—bugs that show up only when the application is heavily loaded,
during stress testing, or even in production, where debugging the application is
most difficult, if even possible.
The ApplicationServlet class is provided in the framework and does not
need to be subclassed to be used in an application. In Tapestry, the application
servlet is just a gateway between the stateless, multithreaded world of the HTTP
protocol and the Servlet API, and the normal, stateful, and single-threaded world
of Java and JavaBeans.
7.2.1 Servlet request processing
You’ve already seen how natural it is to code in Tapestry; components can communicate with each other by setting and reading properties of their containing
page. There’s no awkward shuffling of values in and out of request or session
attributes. Form submissions result in updates to page or domain object properties.
272
CHAPTER 7
Global application data is stored in the Visit object, which persists between
requests, safely stored as an HttpSession attribute.
All of these are examples of the kind of natural, stateful coding that the Servlet API discourages you from using. Tapestry allows you to bypass awkward, stateless coding in two ways: by shifting the coding focus away from the stateless
servlet, and by using object pools.
Object pools are a well-established design pattern used to manage scarce
resources. A common example is a database connection pool, where a small number of database connections are shared by a large number of potential clients.
Objects are checked out of the pool for short periods, to service particular
requests, and then returned to the pool for later reuse. Tapestry uses this pattern
to pool instances of page objects (including all the related objects and components
within a page). This ensures that, while a page (and the hierarchy of components
in the page) is being used to service a request, only a single thread will access the
page. The set of pages for the current request is managed by the request cycle
object, which only returns the objects to the pool at the end of the request.
With Tapestry, the servlet is not the focus of request processing as it is in traditional servlet applications. The servlet is merely a gateway from the Servlet API
to the Tapestry framework. The real work is done by the engine. The engine is
the central hub for processing requests and is particular to a single client session.
It is stored as an HttpSession attribute and so persists between requests. The
servlet’s main job is to find the engine in the HttpSession (or create a new
instance) and invoke the service() method on the instance.
Figure 7.1 shows the processing that the servlet performs on a typical request (a
request where the HttpSession already exists and the engine has already been created). All the magic in Tapestry occurs inside the engine’s service() method. This is
where the request is processed and a reply is constructed and returned to the client.
After executing the service() method, the application servlet stores the
engine instance back in the HttpSession. At first glance, this may not seem necessary—getting the engine instance doesn’t remove it from the HttpSession, so
why is it necessary to store it back in the session? In a word: clustering.
7.2.2 Understanding server-side state
All web applications have some form of server-side state. It may be some transitory state information that exists during a single active request, or it may last
longer, persisting from request to request or session to session. Table 7.2 lists the
four types of server-side state, showing typical approaches using just the Servlet
API and corresponding approaches using Tapestry.
Table 7.2
Types of server-side state
Type
a
273
Servlet API
Tapestry
Temporary variable used
when rendering response
(example: loop index)
• JSP bean with page or request
scope
• JSP local variable (using Java scriptlet)
• HttpServletRequest attribute
• Page or component property
Client-specific data, available in later requests
• HttpSession attribute
• HTTP cookie
• Persistent page or component
property
• Visit object, or property of
Visit object
Non-client-specific data,
available in later requests
• ServletContext attribute
• Static variablea
• Global object, or property of
Global object
Long-term data (outlives any
single user session)
• HTTP cookie
• RDBMS
• HTTP cookie
• RDBMS
Because of the way class loaders work inside an application server, especially in the context of hot application redeployment, making use of static variables is not a recommended approach.
Figure 7.1 The servlet creates the RequestContext façade on the Servlet API objects and then
locates the engine in the session. The service() method does all the actual work of processing
the request and producing a reply. At the end of the request, the engine is stored back in the session.
274
CHAPTER 7
As you’ll see later in this chapter, Tapestry is often using the Servlet API and hiding the details. For example, persistent page properties are stored between
requests as HttpSession attributes.
7.2.3 Managing server-side state in a cluster
The Tapestry framework is designed to support not only highly complex applications but also extremely large applications—applications with thousands of
concurrent users. In the Java 2 Enterprise Edition (J2EE) world, the approach
used to handle extremely high volumes is clustering: dividing the load across
many servers that work together. A collection of these servers is a cluster. To the
outside world, a server cluster appears as a single server, and maintaining this
fiction requires a mix of hardware and software.
Figure 7.2 shows how this works. Clients direct their web browsers to a particular Internet address. The device at that address is not an application server, but
a router or load balancer. It is a specialized server that forwards the request
through a firewall to one of the cluster’s application servers to process the
request. The firewall prevents the client web browser from directly communicating with the application servers (or any other servers within the inner network).
The router and the application server cluster work together to implement a
load-balancing and fail-over strategy that determines which server each request is
sent to.
Figure 7.2 The client web browser connects to a router, which forwards the request through a
firewall to one of the servers in the cluster. If one server fails or becomes overburdened, another
server can take over the load. The client has no knowledge of what goes on “behind” the router.
275
Server clusters can be configured in many ways, according to both the specific
implementations provided by various vendors and the specific needs of an application. The bottom line is that, when an application is executing in a cluster,
there is no guarantee that two successive requests from the same client will be
processed on the same server. Between one request and the next, any server
within the cluster may be shut down, fail unexpectedly, or move processing of
requests to a different server within the cluster because it is overburdened (that
is, for load-balancing reasons). In some configurations, every request is processed by a different server. The client should never be made aware of any of
these transitions. The client sends requests and gets responses and should have
no knowledge of how those requests are satisfied.
When discussing clustering issues, it is easiest to assume that requests are
handled by a primary server until the primary server either fails or deliberately
shifts request processing to a second server (the backup server) due to load balancing. Although this may not match the actual configuration of a cluster, the
relevant issues concerning server-side state are still accurate.
Serializing and copying attributes
One of the prerequisites for any clustering scheme to be successful is that any
server-side state needed for processing the requests that was known on the primary server must be available on the new, backup server. In terms of Tapestry,
the presentation layer, and the Servlet API, this means that any attributes stored
into the HttpSession on the original server must be copied over to the backup
server before it processes any requests. It also means that any server-side state in
the application must be stored as HttpSession attributes, even though Tapestry
frequently represents such state as properties on ordinary objects (which just
happen to be Tapestry pages and components).
When using the Servlet API (with or without Tapestry) in a cluster, any HttpSession attributes are copied from the primary server and broadcast to one or more
backup servers within the cluster. Different application server implementations have
different strategies for clustering, all involving serializing the data (converting objects
into a stream of bytes) stored in the HttpSession on one server and deserializing the
data back into objects on one or more other servers. The serialization may occur
immediately or after some delay. The deserialization may occur constantly or only
just as a backup server is activated to process a request. Despite this, the common factor is that information in one server’s HttpSession is serialized at some point and is
deserialized within a different server in the cluster. Any information stored in the
original HttpSession that can’t be serialized and later deserialized is lost.
276
CHAPTER 7
So, for a servlet application to be a good clustering citizen, it must ensure that
any attributes it stores into the HttpSession are serializable. Because the vast
majority of applications store only simple objects such as String, Integer, Boolean, and Date as their HttpSession attributes, this is not often an issue; all of those
classes are already serializable. Tapestry does the same—the primary object it
stores in the HttpSession is the engine itself, and the engine is serializable.
Keeping synchronized
There is still a remaining gotcha: the question of when the data is copied. This is
left deliberately unspecified in the Servlet API specification. Although various
approaches are possible, Tapestry conforms to the least common denominator
approach: It expects that attributes are copied individually when the setAttribute() method of the HttpSession object is invoked. In fact, this is the exact
approach taken by the popular WebLogic application server.
If the only objects stored as HttpSession attributes are Strings, Integers, and
Dates, then the question of when the copy takes place is not relevant. These
objects are immutable, which means they have no changeable internal state.
However, if an HttpSession attribute is mutable (for example, a List, Map, or custom object), then the question of when serialization and copying takes place
becomes relevant. Figure 7.3 shows how, with mutable HttpSession attribute values, it is possible for the server-side state on the primary server to get out of sync
with the backup server, simply by changing an HttpSession attribute’s internal
state after storing the attribute in the HttpSession.
Once that internal state has been changed, the backup server’s state no
longer matches the actual state stored on the primary server. If the primary
server fails over to the backup server, the backup server will operate with incorrect server-side state, which can lead to incorrect behavior. This class of bugs can
be subtle and confusing, difficult to reproduce, and infuriating to debug in a
production system. Tapestry makes every effort to ensure that this scenario never
happens.
Returning to Tapestry and the ApplicationServlet, the last step shown in figure 7.1 is a call to setAttribute().1 Given that the engine’s internal state is, in
fact, quite mutable, this ensures that any changes to the engine instance are
properly distributed to the other servers in the cluster.
1
Actually, setAttribute() doesn’t occur on every request; some simple optimizations are employed
to avoid invoking setAttribute() when internal state can’t have changed, such as a request that does
not change a persistent page property and does not even access the Visit object.
Understanding the Tapestry engine
277
Figure 7.3 Changing a mutable object’s state after storing it as an HttpSession
attribute can lead to a lack of data synchronization between the primary and backup
servers.
7.3 Understanding the Tapestry engine
At the center of every Tapestry application is an engine, which is the object that
supports and organizes all aspects of the application. The engine binds together
all the smaller subsystems that form a Tapestry application. Its primary concern
is managing server-side state; it manages the Visit object described in chapter 2,
which is where most application-wide state is stored. It also is integral in managing persistent page properties, which allows individual pages to have internal
state that is available from one request to the next.
The most visible aspect of the engine is the service() method, which is the
way in which incoming requests are processed and results are returned to the client
web browser. The service() method is outlined in figure 7.4. This sequence introduces yet another layer of delegation in Tapestry: the engine service, represented
278
CHAPTER 7
Figure 7.4 Execution of the IEngine’s service() method. The engine initializes for the current
request, locates an IEngineService instance for the request, and delegates to the service.
279
by the IEngineService interface. Just as the servlet delegates to the engine to
perform the processing of the request and the rendering of a response, the
engine delegates to an engine service. Engine services, covered in the next section, are the point at which request processing really begins.
Before invoking the engine service, the service() method performs many
initializations, creating and configuring the subsystems within Tapestry. More
important, the engine’s service() method includes multiple levels of exception
catching and reporting. Any uncaught exceptions that percolate back to this
level will be reported using the Exception page, which provides a detailed report
of the exception and nested exceptions as well as all the available Servlet API
objects. An example exception report is shown in chapter 1, figures 1.5 and 1.6.
If the application has become so unstable that even the Exception page fails,
then exception output is written to the console.
7.4 Understanding engine services
In a traditional servlet application, one of the most challenging parts in managing the complexity of an application is in the generation of, and interpretation
of, application URLs. Tapestry handles this problem through engine services.
Engine services are the real link between requests and components.
7.4.1 What’s the problem with application URLs?
Application URLs represent the various operations that can occur inside the application. The most common approach is to have an individual servlet for each operation, and to use query parameters to provide the parameters to the operations.
With small applications, or even initially with large applications, the complexity of managing URL s and query parameters isn’t readily apparent.
There may be some namespace conflicts (everyone on the team will want to
map his or her servlet to /add or /showDetail), but with a little prefixing or suffixing ( /addCustomer, /addAddress, /showProductDetail), these issues can be
ironed out.
A larger risk is exposed as the application grows and changes. Only a tenuous
relationship exists between the JavaServer Pages (JSPs) that construct and render
the application URLs and the servlets that interpret those same URLs. As we’ve
discussed before, the JSPs and the servlets are weakly bound. If a servlet’s mapping changes (say, from /add to /addCustomer as new types of things that can be
added are integrated into the application), every JSP that could possibly reference the servlet must be manually checked, and possibly modified, to ensure that
280
CHAPTER 7
the new correct application URL is being generated. Likewise, any change to the
query parameters—anything from a change in name to a change in what kind of
data is in the parameter—will have a similar ripple effect.
In a nonweb application, you would be changing specific Java methods, and a
source recompile would spot the errors immediately. Modern integrated development environments (IDEs), such as Eclipse, do not even make you wait for a recompile to identify errors; modern IDEs include refactoring tools for automating many
common code changes, such as adding or changing method parameters.
In the loosely coupled world of servlets and JSPs, there is no recompile stage
and no built-in check. Even running the application and visiting all the pages
will not necessarily uncover these mismatch errors. Remember that the servlets
are not ultimately the pages of the application. Servlets are operations and serve
as the transitions between the pages of the application. This means that the only
way to verify the application is to click every link and submit every form that
could possibly be affected by the original change, no matter how minor.
The core of this problem is the separation between the code that constructs
the URL and the servlet code that interprets and acts on the URL. This separation can be addressed with a bit of discipline, by centralizing the URL generation
code. For example, you could create public static methods on the servlet to generate application URLs. This approach ensures that the URL generation and
interpretation code is all in the same place. This approach can be further
improved with a custom JSP tag for each servlet (to produce the URL). However,
any solution that relies on throwing custom code at the problem is suspect. Tapestry takes a different approach to organizing the application.
7.4.2 How does Tapestry handle application operations?
Tapestry is not immune from this issue, even though it contains only a single
servlet. It still has to encode (into the URL) what operation to perform, what
objects are involved, and any additional parameters needed by the operation.
Essentially, Tapestry multitasks its single servlet to accommodate all manner of
application operations, many linked to particular pages and to particular components, some global to the entire application. You’ve seen the three most common of these operations in the examples in previous chapters:
■
Display of a page within the application
■
Response to a link or form submission
■
The default operation that starts an application
281
In Tapestry terms, these operations are provided by engine services. Engine services
are objects that implement the IEngineService interface. This interface defines
methods for both creating and servicing application URLs, all in a single object.
Engine services are much like servlets in implementation. Engine services are
shared by many threads and may hold no client-specific state. Tapestry includes
a default roster of nine services (shown in table 7.3), of which four (home, page,
direct, and external) are commonly used. Most of the services have a corresponding component, such as the page service and the PageLink component, or
the direct service and the DirectLink component.
Table 7.3
Tapestry’s engine services
Service
Description
action
Invokes a listener method indirectly by rewinding the entire page, not just the contents
of a form. Rarely used; the direct service is more efficient.
asset
Dynamically downloads an image or other asset file stored on the classpath (inside a
JAR file). Automatically used with component assets from the framework or from a
component library.
direct
Allows a component to directly respond to a request; used with most links and forms.
external
Allows “bookmarkable” links to particular pages.
home
Default service, used to display the Home page of an application.
page
Responds with a particular page within the application.
reset
Discards all cached data, including all specifications and templates. Used only during
development.
restart
Discards the current session and displays the Home page.
tagsupport
Special service used by the Tapestry JSP tag library to generate URLs for inclusion in a
JSP.
Services are not created equal. Some are more complicated than others, in terms
of what information they encode into a URL (and, correspondingly, what information they later parse out of a URL). Let’s start by looking at the simplest service: the home service.
7.4.3 Using the home service
The home service’s job is to render the Home page of the application. The
application’s Home page is, very simply, the page named Home. You saw this in
chapter 2, where the Home page for the Hangman application was the page displayed when the application was first accessed.
282
CHAPTER 7
Because it does exactly one thing, the home service makes use of no query
parameters. Tapestry assumes that any request with no query parameters, not
even the one to specify a service, is to be processed by the home service. The
home service’s behavior is illustrated in figure 7.5.
In examples in earlier chapters, you saw how a listener method can make use
of the IRequestCycle’s activate() method to select the active page for the current request—the page that will render the response. The home service does the
same, first getting the Home page from the request cycle (using the getPage()
method) and then activating it.
Figure 7.6 illustrates how the activate() method operates: It invokes the
validate() method on the page, which is a general hook provided to pages to
perform basic security validation. This validation allows for application-specific
checks that a page about to be activated is acceptable for the current request and
current client (in effect, the validate() method exists to allow for security
checks—examples of this usage are shown in chapter 10). The page’s validate()
method doesn’t directly implement any checks; instead, pages are allowed to
Figure 7.5
The service gets the Home page from the cycle and activates it.
283
Figure 7.6
The page’s validate()
method invokes
pageValidate() on
any registered
PageValidateListener
objects.
have multiple validation listeners, objects that implement the PageValidateListener interface. The validate() method invokes the pageValidate() method
on each of its listeners.
Most commonly, the page is its own validator. If the page implements the
PageValidateListener interface, then it is automatically registered as its own validator. Individual components within the page may also implement the interface
and be registered automatically as well.
A validator will check to see if the user is allowed to access the page. A validator may adjust the page’s state or, more likely, throw a PageRedirectException
to force a different page to be activated. When a PageRedirectException is
thrown, the processing of the request by the service is aborted. The page specified in the exception is activated and renders a response immediately. Chapter
10 includes an example of a typical use for pageValidate()—to prevent users
from seeing some of the pages of the application until after they have logged in
to the application.
Returning to figure 7.5: after activating (and validating) the Home page, the
home service invokes the engine method renderResponse(). This results in the
active page being rendered and the response being sent back to the client web
browser. Unless a validator has overridden the default behavior, it will be the
Home page that renders the response.
7.4.4 Rendering pages with the page service
Next up the complexity ladder is the page service. This service is responsible for
rendering a particular page, specified by name.
284
CHAPTER 7
Figure 7.7
Page service URL
construction. The
service query
parameter identifies
the service (page) and
the name of the page
to render (Register).
Unlike the home service, which is entirely implicit, the page service has to
encode into its URLs two things: that it is the page service, and which page is to
render the response. All engine services use a pair of standard query parameters
for encoding this kind of information: service and sp.
■
service is used to encode the name of the service, plus any additional con-
text values needed by the service. Multiple values are separated by slashes.
■
sp is a string that encodes a service parameter. The page service doesn’t
use this query parameter, but other services do (such as the direct service).
Figure 7.7 shows how an application URL might appear within the HTML of a
rendered page. Tapestry generates absolute URLs; application URLs always start
with a leading slash and identify the servlet context as well as the application
servlet. Reserved characters, such as the ampersand, will be properly converted
into HTML entities (such as the & in figure 7.9).
The service query parameter starts with the name of the service, page. The
page service has some additional context: the name of the page you want to render. This is also part of the service query parameter. Some services may have four
or five additional context values beyond the name of the service, each value separated by a forward slash.
When the page service is invoked to service a request, it uses the context value
as the name of the page to activate and render the response. Other than that, the
sequence is the same as in figure 7.5, including invoking the page’s validate()
method to perform security checks.
7.4.5 Linking to listener methods with the direct service
The direct service is the true workhorse of Tapestry. The vast majority of links
and forms make use of this service. It is used to trigger an action defined by a
component, either a DirectLink or a Form. A DirectLink component will notify
285
its listener immediately (as described in chapter 2), whereas a Form component
will perform a rewind to process the form submission and then notify its listener
(as described in chapter 3).
Checking for a stale session
One aspect of bridging the stateless HTTP protocol to stateful application development is that the server has to guess when the user has finished using the application. After some period of inactivity, the HttpSession is invalidated. Once it’s
invalidated, any attributes stored in the session are discarded. The server (or
cluster of servers) will no longer have any knowledge of the client, or any serverside state for the client.
Complications can occur if a user steps away from his or her computer for a
period of time and then returns and attempts to continue work. In many cases,
links and forms don’t rely on server-side state, so it is irrelevant whether the state
has been lost between the time the page was rendered and the new request.
In those cases where server-side state is relevant, the framework can validate
that the session still exists. This validation is implemented by the direct service,
which optionally checks for stale sessions. The direct service records into the
URL it generates (when a page renders) whether the application was stateful at
the time the URL was created. When the URL is later processed, the direct service
can check to see if the session still exists or if it has timed out. This shifts the
responsibility for checking for stale sessions away from you and onto the framework—all you have to do is inform Tapestry whether the link or form in question
requires a valid session.
Both the DirectLink and Form components include a stateful parameter
that defaults to false. Setting it to true enables the stale session check. If the session has been invalidated and the component is stateful, then the user is sent to
the StaleSession page. Figure 7.8 shows the default StaleSession page.2
Figure 7.9 shows how the direct service encodes information into the application URL. The direct service has several service context values: the stateful flag,
the name of the page, and the component ID path within the page.
A stateful flag indicates whether the application was stateful at the time the
link was generated. If the value is 1, then the session validation check will take
place when the URL is triggered. Also included in the service parameter value is
the name of the page containing the component, as well as the ID path of the
2
It is possible to override this page in an application, by creating a page in the application named
StaleSession.
286
CHAPTER 7
Figure 7.8 The default StaleSession page, presented when a stateful request is sent after the
server-side session has expired. This page can be easily overridden.
component. In this example, the component “select” on the page “Guess” is referenced. If the component was contained within another component (which is
quite often the case), then the component ID path would be a series of IDs separated by periods.
This URL is an example of a link from the Hangman application in chapter 2,
and it uses a single service parameter to identify which letter the user is guessing. The sp query parameter is used for this purpose. The value, cu, is an encoding of the type (c is for Character; other types have their own prefix) and the
value, the letter u.
Don’t be concerned by all this talk of encodings, URLs, and query parameters. Short of creating your own engine services, you will almost never have to be
Figure 7.9 The direct service URL identifies the page and component and allows for
application-specific service parameters.
287
concerned with these issues. What’s more important are the exact steps taken by
the direct service when processing a request.
Processing a request with the direct service
Figure 7.10 shows the processing the direct service performs for a request. This
is important, because the order in which the operations take place can be critical.
For example, the validate() check occurs very early (as part of the call to the
Figure 7.10 The direct service gets the component, which implements the IDirect interface,
and invokes the trigger() method on it before rendering a response. The trigger() method
invokes the component’s listener method (not shown).
288
CHAPTER 7
activate() method), and won’t have access to the service parameters. The stateful check occurs after validate().
Like the page service, the direct service begins by getting the page (as provided in the context). The validate() method is invoked on the page; then the
component is located within the page. The component must implement the
interface IDirect (which is an extension of IComponent). This interface is implemented by the Form and DirectLink components. After the optional stale session
validation, the service parameters are extracted and stored in the request cycle’s
serviceParameters property, for later access by a listener method. Finally, the
IDirect method trigger() is invoked.
For a DirectLink, the trigger() method will result in a listener method being
invoked, as shown in figure 7.11.
The direct service doesn’t know, or need to know, what the component is. It
only needs to know that it implements the IDirect interface. Forms also use the
direct service and implement the IDirect interface. For a Form, some additional
steps occur; the exact sequence is shown in figure 7.12.
The key aspect of form processing, as covered in chapter 3, is the rewind
phase. When a Form component is triggered by the direct service, it works with
the request cycle to perform the rewind. The page render events (the pageBeginRender() and pageEndRender() methods) occur even when only a form within the
Figure 7.11 The direct service invokes the trigger() method on the DirectLink
component, which obtains its listener and invokes actionTriggered() on it. This results
in the listener method being invoked.
289
Figure 7.12 Form submission using the direct service. The page is notified before
and after the rewind occurs.
page is rewinding. This is very useful; in the examples in chapters 4, 5, and 10,
we make use of this to ensure that empty objects exist for form element components to write values to.
The direct service (with its companion component, DirectLink) is useful for creating a link to an operation within a running application. However, the direct service exposes quite a bit about the internal structure of the application inside its URL.
Inside the context is not just the page name, but also the component ID, or even a
component ID path if the component is contained within another component.
290
CHAPTER 7
Exposing such construction details of a component is acceptable during the
duration of a user’s session with the application because the construction of the
application will not change mid-session. However, the URLs created by the direct
service are not acceptable for long-term storage, such as being recorded as a web
browser bookmark. The application URLs generated by the direct service are
often stateful and, worse, contain elements (the component ID path) that may
change as the application is extended and changed over time.
7.4.6 Creating bookmarkable links using the external service
The external service exists specifically to support URLs that have a long shelf
life; it exposes much less of the structure of the application (just the name of the
page). The external service represents a step back toward a more JSP-like
approach to application development, where certain pages have one specific
purpose. Effectively, the external service is a cross between the page and direct
services. Like the page service, the context is the name of a page. Like the direct
service, service parameters can be passed along in the URL. The processing of
the external service is shown in figure 7.13.
For a page to be the target of the external service, it must implement the
interface IExternalPage, an interface that extends the standard page interface,
IPage. This interface defines a single method, activateExternalPage(), which is
invoked by the service after it activates the page. The external service always
assumes that requests can operate “cold,” without an active session, since the
entire point of the service is to create application URLs that can be bookmarked.
Chapter 10 includes several examples of using the external service in the context
of a real application.
WARNING
If you forget to implement the IExternalPage interface, you’ll get the
following error when you click an external link: Page DatePicker does not
implement the IExternalPage interface.
This service is useful for pages that have a single, specific function. For example,
an e-commerce application can use the external service to create links to product
catalog pages, and will encode product IDs or SKUs as service parameters.
The ExternalLink component is used to create links to pages using the external service. It has a page parameter that specifies the externally addressable page
(like a PageLink component), but it also has a parameters parameter (like a
DirectLink component) used to provide service parameters to the target page.
291
Figure 7.13 The external service validates the page and then invokes the method
activateExternalPage() before rendering the response.
7.5 Page rendering in detail
The point of any web application is to get HTML into the client web browser
where it can be displayed to the user. Perhaps that’s too specific; it may not be
HTML—it may be some other markup language such as XML, XHTML, or WML
(Wireless Markup Language). And the client at the other end may not be a web
browser; it could be a cell phone or personal digital assistant (PDA). Although
Tapestry is primarily focused on HTML, it is designed to operate in any XML related markup language.
292
CHAPTER 7
Rendering in Tapestry is tied to many other subsystems with the framework,
including localization and persistent page properties. At the core of rendering in
Tapestry is the IRender interface, which declares a single method, render():
public void render(IMarkupWriter writer, IRequestCycle cycle);
This interface is implemented by any object that wishes to participate in the
page-rendering process. As shown back in chapter 1, figure 1.11, IRender is a
parent interface of IComponent, so as expected, all components may render. This
interface is simple: The writer parameter will be an instance of IMarkupWriter,
which can be used to render markup output. The cycle parameter is the request
cycle, the object that represents the current request being processed. Through
the request cycle, it is possible to obtain an instance of any page in the application, the application engine, and all the objects of the Servlet API.
7.5.1 Using markup writers
The IMarkupWriter interface is similar to the JDK’s java.io.PrintWriter class.
Although the interface defines a number of overloadings of the print() method
for different parameter types, markup writer instances perform extra work
before writing output. The input is scanned for reserved character values, which
are converted to HTML entities. This means that reserved characters, such as <
and &, are converted to their corresponding HTML entities (< and &,
respectively). It also means that non-ASCII characters and nonprinting characters are converted to numeric entities.
Once again, you see the simplicity and consistency goals of Tapestry; by
using IMarkupWriter, and not simpler interfaces such as PrintWriter or OutputStream , all code throughout the application can easily output safe HTML
(HTML that will display properly in the user’s web browser). In addition, IMarkupWriter has extensions (seen in chapter 5) that streamline the generation of
markup output—special methods for outputting elements and attributes, simply and consistently.
We use HTML as an example here. Different implementations of IMarkupWriter exist for different languages, and the list of reserved characters and entities you can use vary. This means that components don’t have any special
understanding of the content type; that information is isolated inside the
markup writer.
These conversions are useful—and very important. They mean that whether
you’re using components such as Insert or writing output directly within Java
code, you don’t have to constantly check for invalid characters or explicitly run
293
output strings through filters. The filters are built directly into the markup writer
implementation. This affects both usability and security.
End users will never be able to enter input into a form that “breaks” the page.
For example, entering HTML markup as part of the example Register page
(from chapter 5) is perfectly valid. In figure 7.14, the user includes <font size=
"+5"> as part of the first name field. In figure 7.15, the ValidField component
displays that value as the default for the field. With no special checks inside
ValidField, it is the markup writer that scans and corrects the illegal characters.
The text is rendered as <form size="+5">, and the end result is that the
default value in the field matches the user’s actual input.
When the form is corrected and submitted, the output (shown in figure 7.16)
continues to conform to the user’s expectations. The input provided is accepted
Figure 7.14
Entering HTML markup into a TextField could be a problem.
294
CHAPTER 7
Figure 7.15 The markup writer used by Tapestry components, including ValidField, properly escapes
invalid characters that are HTML entities.
and displayed as the output, regardless of the fact that the user entered characters with a specific meaning to HTML. Again, the Insert component used to display the first name can rely on the markup writer to perform the necessary
filtering of HTML characters into HTML entities. If this filtering was not present,
the unwanted HTML tag would not appear, and the formatting of the rest of the
page would be affected, as shown in figure 7.17.
Security can also be affected when user input is reproduced without filtering.
A known exploit used on many online forums is to enter malicious JavaScript
into a text area of a form, including a <script> element. When another user
reads the “message,” the <script> tag and malicious JavaScript are executed
within that user’s web browser. Because of the filtering inside Tapestry, this
295
Figure 7.16 All components get the filter benefit automatically, so users see as output the input
they provided, even though it contains markup.
exploit is not possible; the <script> element entered by the malicious user would
be output as <script>, which would be displayed by the web browser and
not executed.
7.5.2 Going beyond HTML
Much of this book focuses on HTML, but of course the World Wide Web is more
than HTML. Web applications that don’t use HTML but instead use WML (Wireless Markup Language), XML, or XHTML are becoming increasingly useful.
Tapestry is well suited to any kind of HTML or XML style markup. The nature
of any output produced by the framework is controlled by the markup writer
implementation used and by templates of any pages and components.
Tapestry does not validate its page and component templates; that is, it does
not require that the templates conform to any Document Type Definition (DTD),
the way an XML parser would. In fact, the Tapestry template parser is designed
around compatibility with poorly formatted HTML; it is completely accepting of
unquoted attributes and unclosed elements.
The Tapestry template parser finds all the components within the template
by looking for tags with the jwcid attribute. It matches start tags to end tags
and checks that the nesting of dynamic tags is valid. Any text, whether markup
296
CHAPTER 7
Figure 7.17 Without filtering, the way many non-Tapestry applications execute, the user’s input
succeeds in mangling the output of the page, causing text to appear large.
or not, that isn’t part of a dynamic component tag is treated as static, literal
text. This static text will be passed through the client web browser uninterpreted and unchanged. Therefore, in order to generate valid XML, WML, or
XHTML pages, the first step is to ensure that the templates used are themselves valid.
The markup writer implementation provided with Tapestry outputs markup
that is compatible with XML, WML, or XHTML . In other words, the markup
writer properly escapes reserved entities, quotes attributes, and balances start
tags with end tags.
One responsibility for a markup writer is to implement a method, getContentType(), which returns the Multipurpose Internet Mail Extensions (MIME)
type for any content generated with the writer. The only significant difference
between a page that renders HTML and one that renders XML or WML is the
implementation of this method. The default markup writer generates HTML and
297
returns a content type of text/html; charset=UTF-8. The default page class,
BasePage, implements only a single method, getResponseWriter(), which instantiates and returns an HTMLWriter instance.
7.5.3 Understanding the page-rendering sequence
As you’ve seen, once an engine service is ready to render a response, it invokes
the engine method renderReponse(). Figure 7.18 details exactly what happens
inside this method.
The key methods invoked on the page are getResponseWriter() and renderPage(). As described in the previous section, a page indicates what kind of content type it generated by creating an appropriate instance of IMarkupWriter.
IMarkupWriter provides access to the MIME content type it produces with its getContentType() method. This value will be used with the HttpServletResponse
method setContentType().
Rendering a page involves an interaction with the Tapestry persistent page
property subsystem. Once a render begins, it is no longer possible to make
changes to persistent page properties (page properties whose values persist
between requests). Tapestry gives the page one last chance to make changes to
persistent page properties inside the beginResponse() method. After that, it
invokes the request cycle method commitPageChanges(), which notifies the page
recorders (which are responsible for managing persistent page properties) that
they should save any outstanding changes. More details on persistent page properties are coming later in this chapter.
Finally, after all of this maneuvering and setup, the render() method is
invoked. The page will begin rendering the contents of its template, recursively
rendering any enclosed components along the way.
7.5.4 Using page-rendering events
Sometimes a component (or other object) just needs to know when the page is
about to start rendering, or when it has finished rendering. This is commonly
required to set properties of objects that will be used during the render, but only
just in time, as if the page is actually going to render.
These notifications are made available through an event listener interface:
PageRenderListener. This interface include two methods: pageBeginRender()
and pageEndRender().
The pageBeginRender() method will be invoked just before the page’s beginResponse() method. The method is invoked before the page’s recorder is committed, while it is still able to change persistent page properties. The
298
CHAPTER 7
Figure 7.18 The engine’s renderResponse() method. The page is responsible for
creating the markup writer that it, and all components it contains, will use when
rendering. The page’s render() method will render the page and all components it
contains, recursively.
299
pageBeginRender() method will also be invoked before a Form component within
the page rewinds (as discussed in chapter 3).
The pageEndRender() method will be invoked after the page finishes rendering, even if an exception is thrown while the page is rendering. It is also invoked
after a Form component finishes rewinding.
The IPage interface includes methods for adding and removing listeners:
addPageRenderListener() and removePageRenderListener().
As a convenience, if a component implements the PageRenderListener interface, it will automatically be added as a listener to the page that contains it. Pages
may also implement the interface, in which case they will invoke the event methods on themselves. This automatic registration allows for a consistent approach
between pages and components for performing initializations just before the
page renders.
7.6 Loading and pooling pages
The single thing that allows Tapestry to operate so effectively is that Tapestry
applications are built from stateful JavaBeans instead of stateless servlets. This
bridging from the stateless world of HTTP and servlets to the stateful world of
JavaBeans (pages and components are JavaBeans, as are most domain objects) is
both necessary to how Tapestry operates as well as expensive. It is expensive
because a page instance is a complex entity to create. Constructing a page
instance requires a shopping list of work:
■
The page’s specification must be located and parsed.
■
A subclass of the page may be dynamically created.
■
The page is instantiated.
■
Each component of the page must also be located, parsed, extended,
and instantiated.
■
The page’s template is read and applied to the page.
■
Components with templates also must find their templates, and then parse
and apply them.
All told, a single page may contain dozens or more components and hundreds of
supporting objects to represent portions of the template, parameter bindings,
assets, helper beans, and so forth. Despite the fact that specifications and templates are read once and cached for later use, a significant amount of processing
is necessary just to assemble all of these bits and pieces into a functioning page
300
CHAPTER 7
instance that can respond to requests or render a response. Pages are a scarce
commodity that should be used and reused, and not be fruitlessly discarded at
the end of a single request.
However, creating one instance of a page is not enough either. Because pages
(and the components and objects contained within a page) are stateful, they can
only be used by one thread, processing one request, for one user. A servlet, with no
client-specific internal state, can service any number of simultaneous requests in
parallel threads. A page must only be used by a single thread.
As mentioned earlier, the issues that arise when using stateful pages are similar to the issues that arise when using database connections. You can’t afford to
reserve one database connection for each individual user—you could potentially
require thousands of database connections (more than most databases can deal
with effectively), even though only a small number will be in use at any one time.
However, you can’t share database connections haphazardly, because only one
thread can use a connection at any one time. The standard, well-documented
approach to dealing with these issues is the use of an object pool for database connections. The pool contains database connections that are initialized and ready
for use. Code that accesses the database may borrow a connection from the pool
for a short period and then return the connection to the pool as soon as possible.
In some cases, if no connection is available, a new one will be created, on the fly.
Tapestry adopts this same pattern for pages. As a particular page is needed, it is
obtained from a central page pool. If the pool contains no such page (either
because this is the first request involving the page, or because all page instances created thus far are already in use by other threads), then a new page instance is created (along with all of its components, templates, parameter bindings, and other
objects). If a usable page is in the pool, it is removed for the duration of the request.
7.6.1 Retrieving pages from the pool
Once an application is up and running and has processed a few requests,
chances are good that most requests can be satisfied with pages that are already
stored in the page pool. Figure 7.19 shows how the IRequestCycle method
getPage() obtains and readies a page instance when it is needed by a listener or
service. Once a page is obtained in this way, it is cached by the request cycle
object for the duration of the request. Future calls to getPage() with the same
page name return the same instance.3
3
This is a slight simplification. The actual key involves the page name and the desired locale. This is
covered in detail in section 7.9.2.
301
Figure 7.19 The getPage() method of the request cycle. The request cycle goes to the page
source for a ready page instance. The instance is removed from the page pool and attached to the
engine and to the request. The page’s recorder will roll back any persistent page properties to their
prior values.
The request cycle uses the page source to obtain fresh page instances. The page
source is a pool for pages, which has the ability to instantiate a fresh page instance
if none is available in the pool. Pages are attached to the application’s engine via
the attach() method and returned to the request cycle.
The request cycle then gets the page recorder for the page from the engine. The
page recorder is an object that is used to track changes to persistent page properties
for the page. Persistent page properties are properties of pages (or components
within the page) that persist between request cycles. As a persistent page property is
changed, the new value is recorded into the HttpSession for later use (more information on persistent page properties is available later in this chapter).
This is another fundamental aspect of Tapestry: A page’s persistent state, specific to an individual user, is entirely separate from any instance of the page. This
302
CHAPTER 7
is important; it allows us to have our cake and eat it, too—that is, to have a complex object graph without paying the expense of serializing the entire graph to
track changes in the state of objects within the graph.
Java developers are accustomed to thinking that the best way to store the state
of any object is to make the object serializable. Doing this for Tapestry pages and
components would be terribly wasteful. The tiny handful of persistent properties
associated with a page (and its components) are vastly outnumbered by the large
number of supporting objects, most of which should be shared between page
instances. Serializing and deserializing an entire page not only would be a timeconsuming and expensive operation, but also would lead to duplication of common objects that would otherwise be shared.
Tapestry instead concentrates on storing just the persistent page properties,
the few properties that need to be stored in the HttpSession and restored in a
later request. The page recorder object is hooked into pages via a simple notification system, described later in this chapter. The page recorder observes changes
to the persistent properties and safely stores the values in the HttpSession as
named attributes. This approach allows for efficient tracking of changes to persistent page properties, because individual properties are stored as individual
HttpSession attributes. Without the division between page instances and persistent page state, Tapestry couldn’t make any claim to efficiency; with it, Tapestry
can manage complex server-side state simply and effectively.
7.6.2 Creating new page instances
Figure 7.19 covers only the case where a page instance is ready and waiting to be
attached to the request. What happens when there isn’t a ready page instance? In
that case, the page source will create a new instance, using an instance of IPageLoader, as shown in figure 7.20.
The page loader starts by instantiating an instance of the page’s Java class.
Actually, the class that is instantiated will likely be a generated subclass containing the additional fields and methods for any properties specified in the page’s
specification. In any case, first the Java page class is instantiated, and then initial
properties of the page are set, including the page’s name. The page is then
attached to the engine. Next is a recursive process, where each component in the
page is instantiated in a like manner.
As each component is created, the page loader invokes its finishLoad()
method. Components that inherit from the BaseComponent class will load their
templates at this time. Components that contain implicit components in their
template will fire off a new round of component instantiation and loading.
303
Figure 7.20 The page loader’s getPage() method, showing the steps when no pooled page
is available. The page loader creates a new instance of the page and attaches it to the
engine. The page loader recursively instantiates the components of the page. When all
components have been created, it invokes finishLoad() on the page.
Finally, after the entire tree of components rooted at the page is instantiated and
configured, the page’s finishLoad() method is invoked.
One additional operation that occurs in finishLoad() is automatic event registration. When a component implements any of the page event interfaces
(PageRenderListener, PageDetachListener, or PageValidateListener), it will automatically be registered with the page as a listener of that type. A page that implements one of these interfaces will be registered with itself as a listener.
For both pages and components, the finishLoad() method is a good place to
do any final initializations that can’t be expressed in the page or component
specification. Often these initializations involve other components contained with
the page. Once the page’s finishLoad() method completes, the fully initialized
304
CHAPTER 7
and configured page is returned to the request cycle and, from there, to the service or listener method that originally requested an instance of the page.
TIP
Override finishLoad() (in pages or components) to perform final initializations when a page instance is first created.
As a final note, the finishLoad() method has two versions:
public void finishLoad(
IPageLoader loader,
IComponentSpecification specification);
protected void finishLoad();
The first version, the public method with three parameters, is defined in the
IComponent interface. If this method is overridden, the super class implementation must be invoked first.4 The protected version of the method (which takes no
parameters) is a convenience, declared in the class AbstractComponent and
invoked from the public finishLoad() method. The implementation provided by
AbstractComponent is empty and thus doesn’t have to be invoked by a subclass
that overrides it.
TIP
Always override the protected finishLoad() method (the one with no
parameters), unless you specifically need access to one of the parameters in the public finishLoad() method; it’s too easy to forget to invoke
the super class implementation of the public method.
7.6.3 Returning pages to the pool
The flip side of page loading occurs at the end of the request, after the response
has been sent to the client. At this point, the pages that have been attached to
the request must be returned to the pool. More than that, it is necessary that all
the properties of the page (both persistent and transient) be reset to pristine values, ready for use by some later request (for another user in another session).
Leaving property values in place is a dangerous proposition: The same page
instance used by one user in one request will be used by another user in a completely different request. If the first user left his or her credit card number or
password or home phone number in a property and that property is not cleared
4
Failure to invoke the super class implementation will prevent the page or component’s template from
being loaded, as well as cause other side effects.
305
Figure 7.21 The request cycle’s cleanup() method. At the end of the request, each page is
released and stored to the page source’s pool. The detach() method notifies any listeners,
and then invokes initialize(). Finally, the page’s visit, engine, and requestCycle
properties are reset to null.
out, then the next user along may see those values. Figure 7.21 shows the
cleanup sequence.
The page method detach() is responsible for cleaning up the page before it is
stored in the pool for later reuse. The first step is to notify any PageDetachListeners
that are registered with the page. Following that, the initialize() method is
invoked. This method is designed to be overridden in subclasses for this exact purpose: reinitializing any properties. After initialize(), the page’s visit, engine,
and requestCycle properties are set to null. The page should now be devoid of any
changes specific to the user or to the request; the initialize() method should
have reset all such properties, assuming the developers did their work.
306
CHAPTER 7
CAUTION
An application should never store a reference to a page or component
persistently (in the Visit object, or as a persistent page property). Pages should always be self-contained, with only references to other objects
and components within the same page. Instead, the name of the page
should be stored and resolved using the request cycle object. A utility
class, ComponentAddress, exists for this purpose. Storing a page reference will result in one of two unwanted outcomes: Either the page
instance will be illegally accessed while it is in the pool and in an uninitialized state, or the page instance will be accessed by multiple threads.
Both of these scenarios will cause confusing behavior and unpredictable
runtime exceptions.
Having to be concerned with either the detach() or initialize() method is something of a throwback to earlier releases of Tapestry; release 2.3 (the release prior to
3.0) and earlier did not have the <property-specification> element in page and
component specifications. Not only were you responsible for creating the fields
and accessor methods, but you were also expected to implement the initialize()
method to clean up the page at the end of the request cycle. Most applications will
not have to be concerned with page cleanup, as long as they use the <propertyspecification> element for all their persistent and transient properties.
TIP
Use the PageDetachListener interface to perform any cleanups. Implementing the interface ensures that the page will invoke the pageDetached() event method; that way, you won’t have to worry about
invoking super class implementations of methods.
7.7 Using persistent page properties
An important aspect of Tapestry is maintaining the illusion that HTTP is a stateful protocol, and that the user has sole access to the page instances. To the user,
it appears as if the web page in the browser is directly connected to specific
objects in the server. This is truly an illusion—especially when the application is
deployed into a server cluster, since each request may potentially be processed
by a different server, using completely different page instances housed in a completely different JVM.
This same illusion applies to your view of the application as a developer. You
should be able to code a Tapestry application in the same way you would code a
stand-alone application: with the expectation that the object instances continue
to exist from instant to instant. You shouldn’t have to worry about where objects
“go” after your page renders, or after your listener method is executed.
Using persistent page properties
307
Persistent page properties are the key aspect of maintaining this illusion. Each
request is serviced by an equivalent page instance, not the same one, and persistent page properties ensure that equivalence. The persistent properties will be the
same from one request to the next, for as long as the HttpSession is active.
Despite the name, persistent page properties have nothing to do with database persistence. These properties are not stored in a database for long-term
access; they are only available for the duration of the HttpSession. Once the session expires or is invalidated, the persistent properties are lost.
A persistent property is no more than a property whose value is stored as an
HttpSession attribute. When a persistent property changes, the page recorder
observes the change and records the new value for later use. Figure 7.22 shows
how property changes are propagated from the page to the page’s recorder.
A persistent property is only saved persistently as a side effect of invoking the
setter method (which, in turn, invokes the static fireObservedChange() method
of the Tapestry utility class). Refer back to figure 7.13; it identifies a scenario
where the value stored in an HttpSession attribute is not properly replicated. A
similar problem exists with persistent page properties when the property values
Figure 7.22 The implementation of a setter method for a persistent property will invoke to the static
fireObservedChange() method, which notifies the page’s recorder. The page recorder then
constructs a key and stores the value as an HttpSession attribute.
308
CHAPTER 7
themselves are mutable. If, after invoking the accessor method, the property
value is changed, the same lack of synchronization can occur. As a rule of thumb,
you should be careful not to change the internal state of any mutable object after
invoking such an accessor method; otherwise, you will be opening yourself up to
a painful class of bugs when you enable clustering for your application.
Despite the name, persistent properties are not limited to the page. Any component may have persistent properties. The mechanism used is the same,
regardless. Component properties are specific to the page that ultimately contains the component; if the same component is used in two different pages, then
each component will store its individual persistent property as a different
HttpSession attribute. This makes sense from Tapestry’s point of view; these are
different instances of the same component type, not different references to the
same component.
Although it is possible to implement persistent properties yourself, there’s no
reason to do so. Tapestry is fully capable of creating the fields, accessor methods,
notifications, and cleanups necessary to support a persistent property on a page
or a component, using the <property-specification> element in a page or component specification. Appendix D includes complete details about the use and
attributes of the <property-specification> element.
7.8 Using specified properties
A page or component specification may declare entirely new properties for the
page (or component) class. You’ve seen this in previous chapters, where we created several properties in the ToDo application to hold individual ToDoItems and
a list of items.
Tapestry will create an enhanced subclass to support the specified properties.
For components, it may create a subclass even if there are no declared properties, to help support any component parameters. In any case, the enhanced subclass will include new fields and accessor methods. If the property is declared to
be persistent, then the setter method will properly invoke the fireObservedChange() method, as outlined in the previous section.
Declared properties, both persistent and transient, may also have an initial
value. The initial value is either the value of the initial-value attribute of the
<property-specification> element or, if there is no such attribute, the body of
the <property-specification> element. The initial value is an OGNL expression.
The expression is evaluated just once, after the page’s (or component’s) finishLoad() method is invoked. The value for the expression is used to set the property,
309
but it is also saved and used later to update the property when the page is detached
from the request, before being stored back into the page pool for later reuse.
If no initial value is provided for the property, then the framework will read
the property after invoking the finishLoad() method. Again, this value is
retained and used to update the property when the page is detached. This
means that, even for declared properties, it is possible to set an initial value from
the finishLoad() method.
7.9 Localizing Tapestry applications
Localization is the difficult process of creating an application that is translated
appropriately for the user. This means that one user of the application in Boston
may see a different version of the application than a user in Venice. Localization
of an application starts with simple issues, such as translating text into the
proper language, but also encompasses many other potential differences. This
may be represented in small changes, such as the exact way numbers and dates
are formatted, or in large changes, such as the color and layout of entire web
pages. The Java Runtime Environment (JRE) provides a considerable amount of
support for localizing of applications; Tapestry includes well-integrated approaches
for leveraging that support.
Localization of web applications is often an afterthought. Getting web applications working in a single language, typically U.S. English, is daunting enough
that localizing the application is often slipped into “phase 2.” For an application
with broad-ranging appeal, such as a news site, an e-commerce site, or an advertising site, internationalization is a must from the start.
In Tapestry, localization involves two aspects: tracking the desired language
for a particular user and presenting the user interface in that language.5 Often,
the term locale is used instead of language. Locale is a more precise term than
language; it refers not only to a particular language, but also to how that language is used in a particular geographic location. For example, English is a
language, but there are differences between American English and British
English, such as the spelling of some words (color versus colour, for example).
This is resolved by having (in this case) three different locales to represent
English: en, en_US, and en_GB.
5
Localization is often referred to as “l10n” (l, 10 letters, and an n).
310
CHAPTER 7
7.9.1 Using Java localization
In a traditional Java application, the JRE supplies most of the support for handling localization. The first step is to remove hard-coded message strings from
Java code and store them into properties files. A resource bundle is a collection
of such message strings that may be accessed by developer-defined keys.
To handle multiple languages, multiple properties files are used. The base
properties file has a name that ends in .properties; for example, Messages.properties. Additional properties files have the same name, but with a locale name suffix inserted before the .properties extension. These additional files provide
overrides of the keys and values in the base properties file. When you are building up a dictionary of keys and localized values, several of these files can be
involved. The rule is always that more specific values override less specific values. Figure 7.23 shows how individual properties are mixed, matched, and overridden to form localized resource bundles.
The localization for en_US uses just the default values for the two keys: Hello
and color. For the en_GB localization, the value for the color key is overridden to
colour (the British spelling). For the fr localization, both keys are overridden.
Figure 7.23 Keys and values in the main properties file are merged with, and overwritten
by, values from the more specific properties files.
311
So, all of this allows strings to be localized. The next step is to get those strings
used in a web application. Other frameworks, those that use JavaServer Pages,
include JSP tag libraries for accessing a resource bundle and extracting strings
from it. The JSP tags take the place of literal text, with terrible results. The resulting JSP is not even remotely previewable in an HTML editor; it barely looks like
HTML. The loss of literal text in the JSP has an awful impact on developers.
Those little snippets of text act like signposts, helping you to locate the right
form, or the right link, or the right image within a page. Stripping them out
makes every change an agonizing search for the right spot within the page.
As you’ll see, Tapestry takes a more pragmatic and less intrusive approach.
7.9.2 Using Tapestry’s localization features
Tapestry applications allow for localization of both text and images. Localized strings
can be easily referenced in the HTML template, both for inclusion as part of the static
portion of the page and as values bound to component parameters. Images (or really,
any kind of asset) are also automatically localized. Figure 7.24 shows a sequence of
pages demonstrating how Tapestry application localization operates.
After the user selects a new locale (labeled more generally as a Language in the
user interface), the confirmation page is displayed. From there, the user returns to
the original page, but the text, graphics, and even the selections within the dropdown list have changed to reflect the new localization of the application.
Creating a localized HTML template
Listing 7.1 contains the HTML template for the L10N (Localization) page. It
demonstrates two different ways to use localized text in a template: as a value
bound to a component parameter and as a literal value within the template.
Listing 7.1 L10N.html: HTML template for the L10N page
<html jwcid="@Shell" title="message:title">
b
<h1><span key="title"/></h1>
<span key="blurb">Descriptive text.</span>
c
<span key="language-label">Language</span>:
<select jwcid="@PropertySelection"
value="ognl:engine.locale"
model="ognl:localeModel"/>
d
<p/>
312
CHAPTER 7
<input type="image"
src="images/L10N/Change.png"
jwcid="@ImageSubmit"
image="ognl:assets.change"
alt="message:change-label"/>
</form>
e
<p><a href="#"
jwcid="@PageLink"
page="Home"><span key="return-to-home">Return
to Home page</span></a>.
</p>
</body>
</html>
f
Figure 7.24 The user selects a language from the drop-down list and is presented with a
confirmation page in the new language. Returning to the original page shows changes to
text, including the window title bar, as well as images, and even changes to the contents
of the drop-down list.
b
c
d
e
f
313
The title parameter is bound to the localized message for the key title.
The <span> element is replaced by the localized message for the key blurb.
The <span> element is replaced by the localized message for the key language-label.
The alt informal parameter is bound to the localized message for the key
change-label.
The <span> element is replaced by the localized message for the key return-to-home.
A localized string component parameter is used in the very first line of the template:
<html jwcid="@Shell" title="message:title">
Just as the ognl: prefix is used to identify a parameter that is actually an OGNL
expression, the message: prefix is used to identify a parameter that is a localized
message key. The title parameter of the Shell component will be bound to a
specific message string, found by looking up the key title in the localized messages for the page.
NOTE
Components include a getMessage() method for accessing localized
messages for the component. It is possible to replace message:title
with ognl:getMessage("title") to achieve the same result, though the
message: prefix is more efficient, since the framework will know that
the value obtained this way is invariant (for the OGNL expression, it
must keep reevaluating the expression on each use). The getMessage()
method, as part of an OGNL expression, can be used when the localized
message must be combined with other, nonlocalized text to form a parameter value.
These localized messages are stored in a set of properties files for the page. The
properties files are stored in the WEB-INF folder of the web application, alongside the page specification. This example application can be localized in the
default language (U.S. English) as well as French, Spanish, and German; therefore, there are four properties files, named L10N.properties, L10N_fr.properties,
L10N_es.properties, and L10N_de.properties. The first file, L10N.properties,
contains the default language values:
title=Localization Example
blurb=You may change the language used in this application using \
the form below.
language-label=Language
change-label=Change
return-to-home=Return to Home page
314
CHAPTER 7
The other files contain the translations of the values appropriate for the locale.
For example, the French localization, L10N_fr.properties, contains
title=Exemple De Localisation
blurb=Vous pouvez changer la langue utilisée dans \
cette application en utilisant la forme ci-dessous.
language-label=Langue
change-label=Changez
return-to-home=Revenez au Home Page
Having many small message properties files has advantages over more common
approaches that use a single centralized message properties file employed
throughout the entire application. With a single large file, keys often have to be
prefixed to identify which page within an application they belong to. In Tapestry,
short keys are acceptable because each message properties file is used by only
one page or component.
As can be seen in the template, localized message references can be used as
both formal parameters (the title parameter of the Shell component) or informal parameters (the alt parameter of the ImageSubmit component).
A second use for localized messages is to create a localized literal string within
the template, not as a parameter to a component. The instructional text at the
top of the page (“You may change the language …”) is an example of the kind of
literal template text that should be localized. It’s given the localization key blurb
and included in the rendered page by using a specialized <span> element:
<span key="blurb">Descriptive text.</span>
As with a component, the body of the <span> element is always removed and
replaced with the localized message defined by the key attribute. Having text in
the body is, once again, useful when previewing the template in an HTML editor,
even if the text is a meaningless placeholder. This localized literal construct isn’t a
component, but it behaves virtually the same as an Insert component:
<span jwcid="@Insert" value="message:blurb">Descriptive text.</span>
There are two differences between a localized literal string and an Insert component:
■
The element must be <span> (though case does not matter).
■
Additional attributes are always simple literals, even if they have an ognl: prefix.
Like an Insert component, the <span> tag is rendered only if there are additional
parameters. This is useful when the <span> tag references a CSS style. Any unsafe
characters in the localized string (characters such as < and &) will be converted to
HTML entities, as described in section 7.5.1. This can be defeated, if desired, by
315
specifying the value true for the attribute raw. This is recognized by Tapestry and
disables the use of HTML entity filtering, allowing you to place HTML markup
directly in the properties file.
Accessing messages in Java code
Frequently, you will need to make use of localized messages inside your own code,
rather than in a template or specification. You might be constructing a specialized
error message, or a detailed window title. The AbstractComponent base class provides several convenience methods for easily assembling such messages.
The getMessage() method takes a string key parameter. The return value is
the localized message string for that key, from the component’s messages. If no
such message string exists, the key is returned in modified form: uppercase and
enclosed in brackets. This allows your application to continue operating even
with missing message strings, but makes it very noticeable that there’s a problem
to be addressed.
Frequently, you will want to assemble a message from pieces. This is accomplished using the format() method. This method takes a string key and one or
more arguments (the method is overloaded to take one, two, three, or an array of
arguments). For example, your component’s message properties file may contain
the following message:
overdue-reminder=This task was due on {0}.
The {0} is a placeholder for the first argument (arguments are numbered starting with 0). You can include any number of arguments. At runtime, you will combine the message format with additional localized messages and strings. For
example, the following method returns a localized warning message if the
currentDate property is later than the dueDate property:
public String getTaskWarning()
{
Date dueDate = getDueDate();
if (getCurrentDate().compareTo(dueDate) > 0)
{
DateFormat format =
DateFormat.getDateInstance(DateFormat.SHORT, getLocale());
String dueDateString = format.format(dueDate);
return format("overdue-reminder", dueDateString);
}
return null;
}
316
CHAPTER 7
Tapestry uses the Java class java.text.MessageFormat to perform formatting.
Because of this, the previous example could be rewritten to take advantage of
more complicated MessageFormat patterns than simple argument substitution.
MessageFormat can automatically apply different kinds of formatters, if you tell it
the types of arguments. The message format string can be rewritten to indicate
the type and format of the argument:
overdue-reminder=This task was due on {0,date,short}.
Putting this pattern information into the message string removes the need to
obtain the localized DateFormat instance (instead, MessageFormat takes care of
this). The getTaskWarning() method can then be reduced to
public String getTaskWarning()
{
Date dueDate = getDueDate();
if (getCurrentDate().compareTo(dueDate) > 0)
return format("overdue-reminder", dueDate);
return null;
}
Using localized messages in this way is ultimately even easier than putting
together nonlocalized messages as Java code. The awkward natural language
portion is isolated into a properties file, and the Java code is streamlined to just
a call to the format() method, rather than a potentially long series of string concatenations. There’s that Tapestry vision again: Creating uniform, localized messages is easier than the quick-and-dirty approach.
Localizing images
Very often, images used in a page contain text that must be localized. In this
example, the ImageSubmit component uses a localized image. In figure 7.24,
the image is initially the word Change, but when the page is later redisplayed in
French, the button is labeled Changez. Since this is an image, it involves more
than just a change in a message: There are two (or more) versions of the same
button, and Tapestry must create a reference to the correct version when rendering the page’s HTML.
Tapestry accomplishes this by using an asset to represent the image. You first
saw assets in chapter 2, when we used this feature to map from logical names for
images to specific files in the web application’s context folder. That is just one
aspect of assets; another is related to localization. Each asset defined in the
page specification may be matched against multiple files, named with the
317
Figure 7.25 A single asset may be mapped to different files, each providing the correctly
translated text.
message strings properties files. Figure 7.25 shows the four different files for the
change asset.
At runtime, the page’s locale is used by the asset to determine the correct
URL to include in the rendered page. When there isn’t a match on locale (for
instance, if the locale was somehow changed to Hutu), then the best default (the
file Change.png) is used.
Editing the engine’s locale
Changing the locale for an application is as simple as changing the engine’s
locale property. The default value for the engine locale is the server’s locale,
which is often U.S. English. When the locale is changed, an HTTP cookie is
recorded in the client web browser so that, in future sessions, the newly selected
locale will be the default.
The engine’s locale is used for a single purpose: when obtaining a page
instance from the pool (or when creating a fresh page instance). Page instances
are stored in the pool using a key; that key incorporates both the name of the
page and the locale of the page. This shows up in figure 7.19, where the IRequestCycle invokes the getPage() method on the page source. The page source
takes into account the engine’s current locale when finding or creating a page
instance to attach to the request.
This all means that whenever a page is first accessed during a request, the
locale for the page will match the engine’s locale. So, if the engine is set to locale
fr, the page will render in French and will use French page and component templates, French assets, and so forth. Once a page instance is attached to a request,
it is used for the duration of that request. Even if the engine’s locale is changed
to a different value in a listener method, the page instance will still be French.
There’s no way to detach a page instance from the request once it’s attached, so
there’s no way to change the locale of an active page. Therefore, after changing
the locale, it is important to load a new page in the new locale.
This shows up in the listener method for the form on the L10N page:
318
CHAPTER 7
{
cycle.activate("L10NResult");
}
By the time this listener method is invoked, the engine’s instance will already
have been changed by the PropertySelection component. Redisplaying the current page (L10N) will, confusingly, show the page in the previous locale. By
instead loading, activating, and rendering a new page, the response will be in
the correct localization.
Localizing the drop-down list
The drop-down list used to select the new language is also localized, in two ways:
■
The labels are localized to the current locale.
■
The selectable values omit the current locale.
This is accomplished by creating an IPropertySelectionModel implementation
for selecting locales. The model is provided by the L10N page, in its Java class:
private IPropertySelectionModel _localeModel;
public IPropertySelectionModel getLocaleModel()
{
if (_localeModel == null)
_localeModel = new LocaleModel(getLocale());
return _localeModel;
}
The first time this method is invoked, it creates an instance of the model, passing
the current locale to the model’s constructor. Note that the page is not required
to release this cached model instance at the end of the request cycle. This property is not specific to the client in any way; it is specific to the locale. Even if the
page did release this model, it would create an identical instance the next time
the method is invoked.
The implementation of the model is shown in listing 7.2.
Listing 7.2 LocaleModel.java: property selection model class
package examples.l10n;
import java.util.ArrayList;
import java.util.Locale;
import org.apache.tapestry.form.IPropertySelectionModel;
public class LocaleModel implements IPropertySelectionModel
{
private Locale _activeLocale;
private List _locales = new ArrayList();
private static final Locale[] AVAILABLE_LOCALES =
{ Locale.ENGLISH, Locale.FRENCH, new Locale("es"),
Locale.GERMAN };
public LocaleModel(Locale activeLocale)
{
_activeLocale = activeLocale;
String activeLanguage = activeLocale.getLanguage();
for (int i = 0; i < AVAILABLE_LOCALES.length; i++)
{
if (AVAILABLE_LOCALES[i].getLanguage().
equals(activeLanguage))
continue;
c
_locales.add(AVAILABLE_LOCALES[i]);
}
}
public int getOptionCount()
{
return _locales.size();
}
d
public Object getOption(int index)
{
return _locales.get(index);
}
public String getLabel(int index)
{
Locale l = (Locale) _locales.get(index);
return l.getDisplayLanguage(_activeLocale);
e
}
public String getValue(int index)
{
return Integer.toString(index);
}
f
b
319
320
CHAPTER 7
public Object translateValue(String value)
{
int index = Integer.parseInt(value);
return getOption(index);
}
}
b
c
d
e
f
This defines the four different locales that may be displayed in the drop-down list.
The active language is removed from the list, leaving just the other three locales.
This should return 3, unless the initial locale is not one of the expected four.
The Locale class can create a localized description.
We encode the index into the list of locales as the client-side value.
The model starts with a list of available locales but excludes the current locale
(matching on language). The rest of the model is similar to those shown in chapter 4, where the desired locale is encoded as an index number into a list of
locales. Locales are self-describing, and the method getDisplayLanguage() is
perfect for our needs because it provides a localized, user-presentable name for a
locale translated into another locale.
Localizing HTML templates
Another option exists for handling localization, one that is applicable to pages
that contain a large amount of text and a small number of components. Rather
than isolating localized text into properties files and then using localized literals
to pull them out, it is possible to provide multiple templates instead.
Tapestry localizes templates in much the same way it localizes assets; it
searches for the best match based on the current locale when the page is constructed. It is not a requirement that the different templates differ only in terms
of visible text. Different localizations of the template may be considerably different in layout and construction, and may even contain a different number of components. Each localization of a page is independent.6
For components that are the same regardless of localization, it is best to use
declared components and put the type and parameters of such components into
6
It is acceptable to vary the implicit components in different localizations of the template. Omitting a
declared component from a localization of the template will result in a runtime warning (on the console), not a runtime error, but should still be avoided.
Summary
321
the page specification. Page specifications are not localized; only the templates
are. Using declared components instead of implicit components reduces the
amount of duplication in the templates—which is a very good thing when you’re
fixing bugs in the page.
Of course, as the developer, you are free to mix and match solutions. You can
certainly use localized literals in concert with localized templates. In addition,
you can use components with their own templates, and they, too, can use whatever mixture of localized resources is best for them.
7.10 Summary
In this chapter we’ve examined some of the problems related to grafting a stateful application on top of a stateless protocol, HTTP. Tapestry uses a variety of
techniques to maintain the illusion that the there is a direct connection between
the web pages viewed in the client’s web browser and specific page objects within
the server. It extends this illusion to you, the developer, as well, allowing you to
code your pages without concern for all the issues related to page pooling and
clustering. The framework imposes only a few modest constraints on you to fit
into this system, and provides declared properties to make it easy to conform to
those constraints.
Tapestry includes many hooks for extending the behavior of individual pages
and components. The pageValidate() method allows for simple security checks,
and the finishLoad() method allows pages and components to perform initializations that can’t be expressed in a page or component specification. By understanding how the different major interfaces (IRequestCycle, IEngine, IPageSource,
and others) work together, you can replace some or all of them to radically
change the behavior of Tapestry.
Integrated into all aspects of Tapestry is the concept of localization. The
framework makes localizing an application as simple as providing translations of
templates, string properties files, and image asset files.
You’ve also seen how Tapestry utilizes engine services to bridge between the
application and the Servlet API. In the next chapter, you’ll see how to combine
the ideas from this chapter, and from chapter 6, with a few new ideas in order to
create components of even greater complexity and reusability.
Advanced techniques
This chapter covers
■
Creating new engine services
■
Generating dynamic client-side JavaScript
■
Integrating Tapestry with JSPs
322
323
In previous chapters, we described the basics of creating reusable components
and gave you an overview of how Tapestry operates internally. Quite a bit can be
accomplished by simply combining existing components and leveraging the
existing services of the Tapestry framework, but even more can be done by
expanding both the server- and client-side processing of components.
In this chapter, we’ll cover three additional advanced Tapestry techniques
and illustrate them with practical examples:
■
■
■
Defining new engine services—You can create new engine services to create new kinds of interactions between the client web browser and the
server. As an example, we’ll create a simple banner ad system.
Generating JavaScript dynamically—Tapestry handles the dynamic generation of client-side JavaScript to allow for significant client-side processing. You’ll see how this works by creating a reusable a credit card
input field component.
Integrating Tapestry with JSPs—You may not always have the freedom to
implement exclusively in Tapestry; fortunately, it’s reasonable to mix and
match a Tapestry application with a more traditional application implemented using servlets and JSPs.
8.1 Creating new engine services
As we discussed in chapter 7, engine services fill the role of servlets within a Tapestry application. Because Tapestry applications are highly structured (in terms
of a component hierarchy within each page), a small number of services are sufficient to implement most functionality.
Most of the engine services (notably the page, direct, and external services)
fall into the same mold:
■
■
■
A page name, and perhaps a component ID, is encoded into the URL.
The page is loaded, and methods are invoked on the page or component.
The active page renders a response.
Not all interactions between the client and the server fall into this mold. For
example, in chapter 6 we described how components may be packaged with private assets, and how the asset service is involved in exposing those assets, stored
on the classpath inside a JAR file, so that they may be accessed by the client web
browser. For the asset service, the request cycle doesn’t involve pages or components at all but only the path to a classpath resource, which is sent back to the client web browser as a stream of bytes.
324
CHAPTER 8
Advanced techniques
New services can be defined easily and are free form. You can create pageoriented services similar to the framework’s page and external services as well as
component-oriented services similar to the framework’s direct service. You can
also create entirely new forms of services that don’t even send an HTML response,
such as the asset service.
To demonstrate how to build these kinds of interactions, let’s create a new
type of service that does not involve rendering a response page.
8.1.1 Defining a banner ad system
A ubiquitous presence on many high-volume web sites is the banner ad. Banner
ads are small images that span the top, bottom, or sides of a web page. Clicking
a banner ad sends the user to an advertiser’s web site. Banner ads are one of the
oldest forms of advertising on the Internet. For this example, we’ve created banners for several popular Java and open-source software web sites. Figure 8.1
shows an example of the demo application in operation. You can try it yourself
by opening a web browser to http://localhost:8080/ads/app.
Figure 8.1 The banner ad is selected at random and includes a tooltip. Clicking on the ad opens a
new window showing the advertiser’s page (but will also record the click-through).
325
It isn’t enough to have the banner ad be a simple link to the target web site.
Advertisers pay popular web sites to display advertising banners. They also need
to know how often users are enticed into clicking on the ad banner. A banner ad
system must track how often a particular banner is displayed (called an impression) and how many times users have “clicked through” from the web site serving
up the banner ad to the advertiser’s web site. The demo application includes a
statistics page for viewing this information, as shown in figure 8.2.
A proper banner ad system as a Tapestry component library will need to meet
the following requirements:
■
Tracks which ads have been displayed
■
Tracks which ads have been clicked through
■
Easily allows new ads to be defined
We could build a banner ad system using the DirectLink component easily enough,
but such a solution would have some problems:
Figure 8.2
The number of impressions and click-throughs for each advertiser is available.
326
CHAPTER 8
Advanced techniques
■
The link generated by the DirectLink component would not be bookmarkable, since URLs created by the underlying direct service contain details
about the structure of the application, which can change over time.
■
The link would work only if the page’s PageValidateListeners do not
interfere, which is not acceptable; clicking through a banner ad should
operate regardless of anything else on the page.
■
A DirectLink would require that a page instance be obtained and persistent
page properties be restored. This is unnecessary overhead for a banner ad
redirect, which never uses or changes page state.
Basically, servicing a request for a banner ad click-through is quite different from servicing an ordinary component-oriented request. We need to bypass page validation,
and instead of activating application logic and responding with a page from the
application, we’ll always be sending a redirect response to the client web browser.
Either of these requirements would lead us to consider creating a new engine service.
To build this system, we’ll take the following steps:
1
Define the data model, including the interface for the banner ad and the
interface for the source of the banner ads.
2
Use an application extension to access the banner ad data.
3
Implement the BannerAd component.
4
Implement the new banner ad service.
5
Create a library to contain the component and service.
Finally, we’ll build a small demonstration application to test the banner ad system.
8.1.2 Defining the data model
The first step is to define a data model for banner ads. This is accomplished in
two parts. First, we need an interface to represent an individual banner ad, providing the following read-only properties:
■
The URL for the banner ad image
■
The width and height of the image
■
The URL of advertiser’s web site
■
A title describing the advertiser
■
A unique ID for the banner ad
327
Figure 8.3
Two interfaces define the
banners, as well as the
source of the banners.
The banner source is also
responsible for recording
impressions and clickthroughs.
In addition, we need an interface to define the source of the banner ads. Its primary job is to provide a random banner ad for inclusion on a page. A secondary
job is to record impressions and click-throughs of specific ads.
Figure 8.3 shows these interfaces, as well as the two implementations.
At runtime, the banner source is used to retrieve a random banner for inclusion on a page (and record the banner as an impression). Later, if the user clicks
the banner, the banner source is used to record the click-through, before a redirect response is sent back to the client web browser to redirect the web browser to
the advertiser’s web site.
The implementations provided, Banner and XMLBannerSource, are deliberately
simple. The XMLBannerSource reads an XML file that contains a list of banner ad
definitions, which are converted into instances of Banner. The Banner class is
largely a data object, providing access to the properties we’ve listed. In addition,
in this simple implementation, the Banner class is used to record the impression
and click-through count.
Obviously, a real production system would be more involved. It would more
likely operate using a database to define the advertisements, and it would store
328
CHAPTER 8
Advanced techniques
impressions and click-throughs persistently back into the database. A real system
would support various types of banners (such as Java applets, embedded Flash
movies, or whatever technology comes next), and the selection of which banner
to show to which client wouldn’t necessarily be random—there would be a way to
target specific ads at particular users.
The first hurdle to clear is coming up with a way to access the banner ad data
provided by the XMLBannerSource instance.
8.1.3 Accessing the data model as an application extension
As you’ll see, two different sections of code need to access the banner ad data.
The BannerAd component will have to access the banner source to get a random
banner to include on the page. If the user clicks that banner, the banner engine
service will be invoked, which will need to again access the banner ad data, in
order to get the advertiser’s web site URL as well as record the click-through.
The design of this simple banner ad system uses only a single list of potential
banner ads, stored as a file within the web application. Ideally, the banner data
should be a singleton object, created as needed by the component or by the service. The banner data singleton should be created just once, initialize itself from
the banner data file once, and be accessible by any number of simultaneous
requests. Tapestry includes a mechanism for just these kinds of singletons: application extensions.
Application extensions are very much like helper beans: They are declared as
part of an XML specification, they are instantiated only as needed, and they can
have their properties initialized. Unlike helper beans, they are not associated
with any single page but with the overall application. Additionally, they are
always singletons.
Application extensions are declared in the application specification. The
specification for the BannerAd demo application includes the following:
<extension name="bannerads.banner-source"
class="bannerads.library.impl.XMLBannerSource"/>
Application extension names are allowed to contain periods and dashes. In
order to avoid naming conflicts, prefix the names with a Java package name.
You can retrieve application extensions by their name. The first time an
extension is accessed, the Java class for the extension is instantiated and configured. This object is stored for later access as a singleton—the same object will
always be returned in any future lookup with the same extension name. This is
necessary in our example because both the BannerAd component (described
329
next) and the banner ad engine service (described in section 8.1.5) must access a
single, shared instance.
8.1.4 Implementing the BannerAd component
A BannerAd component will render an <a> hyperlink element enclosing an
<img> element. The href attribute for the hyperlink will reference the banner ad
service, which will record the click-through before sending a client redirect
response to the web browser. The src attribute for the <img> element will be
dynamically determined from the underlying Banner object.
BannerAd allows informal parameters, which are included as additional
parameters for the <a> element. This allows the page to specify a CSS class
attribute for the component.
The BannerAd component has no template; what HTML it produces is rendered in code. Its component specification, shown in listing 8.1, is also quite short.
Listing 8.1 BannerAd.jwc: specification for the BannerAd component
class="bannerads.library.components.BannerAd"
allow-body="no"
allow-informal-parameters="yes">
<reserved-parameter name="href"/>
The BannerAd component does not allow a body. Although it allows informal
parameters, it reserves the href parameter, since it will render a value for the
href attribute in Java code. The specification for the BannerAd component is
stored on the classpath as /bannerads/library/BannerAd.jwc. Listing 8.2 shows
the BannerAd class.
Listing 8.2 BannerAd.java: Java class for the BannerAd component
package bannerads.library.components;
import org.apache.tapestry.AbstractComponent;
import org.apache.tapestry.IEngine;
330
CHAPTER 8
Advanced techniques
import
import
import
import
import
org.apache.tapestry.engine.IEngineService;
org.apache.tapestry.spec.IApplicationSpecification;
import bannerads.library.BannerService;
import bannerads.library.IBanner;
import bannerads.library.IBannerSource;
public class BannerAd extends AbstractComponent
{
{
Skips render when
form rewinds
return;
IBanner banner = getRandomBanner(cycle);
IEngineService service =
cycle.getEngine().getService(BannerService.SERVICE_NAME);
ILink link = service.getLink(cycle, this,
new Object[] { banner.getId()});
Constructs link to
banner ad service
writer.begin("a");
writer.attribute("href", link.getURL());
Converts link
object to URL
writer.attribute("src", banner.getImageURL());
writer.attribute("width", banner.getWidth());
writer.attribute("height", banner.getHeight());
writer.attribute("alt", banner.getTitle());
writer.attribute("border", 0);
writer.end();
}
private IBanner getRandomBanner(IRequestCycle cycle)
{
IEngine engine = cycle.getEngine();
IApplicationSpecification specification =
engine.getSpecification();
IBannerSource source =
(IBannerSource) specification.getExtension(
IBannerSource.BANNER_SOURCE_EXTENSION_NAME,
IBannerSource.class);
source.initialize(cycle);
Obtains banner
source application
extension
Initializes the
banner source
331
IBanner result = source.getRandomBanner();
source.recordImpression(result.getId());
return result;
}
}
The renderComponent() method invokes getRandomBanner() to obtain a random
instance of IBanner from the banner source. Once it has the banner, the component can then access the banner service and use it to create a link (much like
the DirectArea component in chapter 6, but with a different service). The banner service expects a single service parameter: the ID of the banner that is
being rendered.
The remainder of the renderComponent() method uses the link object
returned from the service, and the banner information provided by the banner
source, to generate an <a> hyperlink element enclosing an <img> element. Any
informal parameters are rendered as part of the <a> element.
One special case to watch out for when creating components involves forms.
When a form is submitted, the Form component will rewind, rerunning the render process for all components it encloses—even components that have nothing
to do with the form itself, such as BannerAd. It would not be correct to count a
form rewind as a new banner impression—so we check the request cycle first
thing to see if a form is rewinding, and do no work in that case.
The getRandomBanner() method is also straightforward. The application specification is obtained from the engine instance; the getExtension() method
accesses an extension, instantiates it if necessary, and returns the shared
instance. In addition, it checks that the extension instance is assignable to the
type passed in as the second argument (that is, the method checks that the
extension either implements the interface or is a subclass of the indicated class).
In this class, the extension must implement the IBannerSource interface.
WARNING
Only invoke getExtension() for an extension that is defined in the application extension. If the application specification does not have an
<extension> for the given name, then the framework will throw an exception: No extension named ‘bannerads.banner-source’ exists in this namespace.
If an extension is optional, then you must use the checkExtension()
method to see if a named extension exists.
332
CHAPTER 8
Advanced techniques
Obtaining the shared banner source instance is not quite enough: We must
ensure that the source is initialized. There’s a wide range of possible implementations of the banner source. The example implementation reads an XML file,
but you can easily imagine a banner source that works with a database, or an
Enterprise Java Bean (EJB), or a Java Messaging Services (JMS) queue, to obtain
banners and record impressions and click-throughs. The initialize() method
is passed the request cycle, which gives the banner source access to the entire
Tapestry and Servlet APIs; this should be sufficient for any banner source implementation to initialize itself.
In addition, we must invoke initialize() every time we get the banner source,
since there’s no way to determine if it was created just then or if it was an alreadyexisting instance.
8.1.5 Implementing the banner service
Implementing an engine service is much like implementing a servlet: Engine
services are shared instances that must be thread safe (they may be accessed by
many threads simultaneously). Unlike a servlet, an engine service is responsible
for building the URL s for the requests it may later service. The getLink()
method builds URLs, and the service() method reacts when those same URLs
are triggered by a request from the client web browser. Listing 8.3 provides the
code for the BannerService class.
Listing 8.3 BannerService.java: Java class for the banner service
package bannerads.library;
import
import
import
import
import
import
import
import
org.apache.tapestry.IComponent;
org.apache.tapestry.RedirectException;
org.apache.tapestry.engine.AbstractService;
org.apache.tapestry.engine.IEngineServiceView;
org.apache.tapestry.request.ResponseOutputStream;
org.apache.tapestry.spec.IApplicationSpecification;
public class BannerService extends AbstractService
{
public static final String SERVICE_NAME = "bannerads.banner";
public ILink getLink(IRequestCycle cycle, IComponent component,
Object[] parameters)
{
return constructLink(cycle, SERVICE_NAME,
333
null, parameters, false);
}
public void service(
IEngineServiceView engine,
ResponseOutputStream output)
throws ServletException, IOException
{
IApplicationSpecification specification =
engine.getSpecification();
IBannerSource source =
(IBannerSource) specification.getExtension(
IBannerSource.BANNER_SOURCE_EXTENSION_NAME,
IBannerSource.class);
source.initialize(cycle);
Object[] parameters = getParameters(cycle);
String bannerId = (String) parameters[0];
IBanner banner = source.getBanner(bannerId);
source.recordClickThru(bannerId);
throw new RedirectException(
banner.getClickThruURL());
Get the banner
ID from the URL
Obtain the
matching banner
Send redirect
to client
}
public String getName()
{
return SERVICE_NAME;
}
}
In Tapestry, application URLs are not represented as strings; they are represented as instances of the interface ILink. You can see this in the return value of
the getLink() method. An ILink is a combination of the URL for accessing the
application servlet and any query parameters (the service and sp parameters,
generally) that must be passed in the request. This distinction allows the link
components (such as DirectLink and PageLink), which use the HTTP GET
method, to encode the query parameters within the URL directly, but allows the
Form component, which uses the HTTP POST method, to encode the parameters
as hidden form fields.
334
CHAPTER 8
Advanced techniques
AbstractService provides a constructLink() method. The last two parame-
ters are an array of strings (that form the service context) and an array of objects,
the service parameters. As you’ve seen with the page and direct services, the service context can be used to identify a page or a component within a page.
Because those concepts are not relevant to this service, we simply pass null.
This service does use a single parameter, the banner ID, which was passed in
from the BannerAd component. The last parameter to constructLink() is a flag
indicating whether the URL should be encoded using HttpServletRequest.encodeURL(). The encodeURL() method encodes the HttpSession ID into the URL, which
ensures that the connection between the client web browser and the HttpSession
on the server is maintained, even when the browser has disabled HTTP cookies
(the normal way of communicating the HttpSession ID from the client to the
server). For most services, the encode flag should be true, but this service doesn’t
require encoding, because triggering a banner ad does not affect any clientspecific state that may be stored in the HttpSession.
TIP
When in doubt about whether to encode or not encode the URL, just
pass true and let the URL be encoded. It costs virtually nothing and ensures you won’t get any nasty surprises from clients who have disabled
HTTP cookies.
Servicing a request
When the user clicks on the banner ad, a new request to the server will be
directed at the banner engine service, which will invoke the service() method.
This method must
■
Obtain an instance of the banner source
■
Ensure that the source is initialized
■
Get the correct banner from the source
■
Record the click-through for the banner
■
Redirect the client’s web browser to the advertiser’s web site
Obtaining the banner source and initializing it is the same here as in the BannerAd
component. It may appear that there is no need to invoke the initialize()
method, because the BannerAd will already have initialized the banner source.
However, this is not the case; the call to initialize() is necessary:
■
The URL to the banner service may have been bookmarked by the user, in
which case there’s no assurance that a BannerAd component has yet executed
335
its renderComponent() method and initialized the banner source (the server
receiving the request may have restarted in the meantime).
■
In a cluster, it is possible that the banner source was initialized within one
server but that the banner service request was received by a different server
in the cluster.
The ID of the banner was provided as a service parameter; the getParameters()
method (provided by the AbstractService base class) will extract the service
parameters from the sp query parameter, and convert them from encoded strings
back into objects. We can then cast the first parameter back into a string:
Object[] parameters = getParameters(cycle);
String bannerId = (String) parameters[0];
Armed with the banner ID , we can get the banner instance from the banner
source, and record that a click-through for that banner has occurred:
IBanner banner = source.getBanner(bannerId);
source.recordClickThru(bannerId);
The final step is to send a redirect back to the client web browser. This is accomplished by throwing a RedirectException. The RedirectException is constructed
with the advertiser’s web site URL:
throw new RedirectException(banner.getClickThruURL());
Note that this is a distinct exception from the PageRedirectException that may
be thrown in a pageValidate() event method. PageRedirectException is handled
internally by Tapestry and simply represents a change in the active page (the
page that will render the response). Throwing a RedirectException here sends
an HTTP redirect back to the client web browser, forcing it to submit a new GET
request to the indicated URL.
Providing the service name
The last requirement of an engine service is that it knows its own name. The service must implement a getName() method to provide this name. This method
ensures that the service name declared in the library specification (discussed
shortly) matches the name used by the service to create links.
The standard procedure is to define a public static field with the engine service name, which not only ensures consistency but (as in the BannerAd component) makes it easier for other classes to obtain the service instance:
336
CHAPTER 8
Advanced techniques
Figure 8.4 Tapestry checks that each engine service matches up against its declared names; if not,
it doesn’t even try to service the request.
public static final String SERVICE_NAME = "bannerads.banner";
public String getName()
{
return SERVICE_NAME;
}
Providing the correct service name is critical to Tapestry. On the first request, it
will instantiate all the services and verify that each service has the correct name.
On any mismatch, it will abort the request (by throwing a ServletException), and
the user will see an Error 500 page. Figure 8.4 is an example of what you’ll see if
there’s a typo in your library specification.
At this point, we have the component and matching service, plus the IBanner
and IBannerSource interfaces (and implementations of those interfaces). It’s time
to package all of this together as a reusable component library.
Before we can use this service and component in an application, we must create a
library to contain them. The library specification, shown in listing 8.4, defines
the service provided with the library. The file is stored in the classpath as /bannerads/library/BannerAds.library.
337
Listing 8.4 BannerAds.library: specification for the banner ad library
<!DOCTYPE library-specification PUBLIC
<library-specification>
<service name="bannerads.banner"
class="bannerads.library.BannerService"/>
</library-specification>
Finally, you see how the banner service is plugged into Tapestry: The <service>
element provides a name for the service and a class to instantiate. As we just discussed, the name must be unique (which is why it is prefixed by a package name),
and the actual service instance must know its name, to ensure that it will generate usable URLs. The BannerAd component is part of the library, because its
specification file, BannerAd.jwc, is also stored in the /bannerads/library folder.
After packaging the classes, interfaces, and specifications together as a JAR
file, we are ready to use this library in an application.
8.1.7 Building a banner ad application
A small demonstration application that makes use of the BannerAds library and
the BannerAd component is included with the book examples, and figures 8.1
and 8.2 show this application in operation. You can try it yourself by directing
your web browser to http://localhost:8080/ads/app.
Using the library is nearly as simple as the examples in chapter 6. The only
extra step is to define an application extension for the banner source. Listing 8.5
shows how the application extension is declared.
Listing 8.5 ads.application: specification for the banner ad demo application
<application name="BannerAd demonstration">
<extension name="bannerads.banner-source"
class="bannerads.library.impl.XMLBannerSource"/>
338
CHAPTER 8
Advanced techniques
<library id="banner"
specification-path="/bannerads/library/BannerAds.library"/>
</application>
The extension creates an instance of XMLBannnerSource, an implementation of
the IBannerSource interface that gets its list of ad banners from an XML file
stored inside the WEB-INF folder. The file, banners.xml, provides the banner
information for four Java and open-source web sites. Listing 8.6 shows the contents of this file.
Listing 8.6 banners.xml: list of banner ads used in the demo application
<banner-ads>
<banner id="0001" title="Cafe Au Lait"
width="468" height="82" imageURL="ads/cafeaulait.png"
clickThruURL="http://www.ibiblio.org/javafaq/"/>
<banner id="0002" title="JavaLobby.org"
width="468" height="60"
imageURL="ads/javalobby.png"
clickThruURL="http://www.javalobby.org"/>
<banner id="0003" title="Apache Jakarta Project"
imageURL="ads/jakarta.png"
clickThruURL="http://jakarta.apache.org"/>
<banner id="0004" title="TheServerSide.com"
imageURL="ads/theserverside.png"
clickThruURL="http://www.theserverside.com"/>
</banner-ads>
The BannerAd component is used just on the template of the application’s
Home page:
<a href="#" target="_new" jwcid="@banner:BannerAd">[Banner Ad]</a>
And that really is all it takes! Using target="_new" directs the client web browser
to open the URL in a new window. We’ve defined a whole new service that handles requests from the client’s web browser in an entirely new way but hidden
that service behind a component. Despite this component’s very special request
339
processing (recording the click-through event and sending a client redirect to a
completely different web site), no requirements were forced on the page or on
the application containing the page. The BannerAd component can be literally
dropped right into a page without concern for how it operates internally, and it
can be added to any application merely by providing a banner source in the
application specification. This level of automatic integration is a key distinguishing feature of Tapestry, something that is simply not possible when using ordinary servlets and JSPs, or even prevailing frameworks such as Struts.
The banner ad examples illustrate how flexible Tapestry is in terms of serverside request processing. As you’ll see in the next section, Tapestry has an equally
in-depth approach to client-side scripting.
8.2 Client-side scripting
An unavoidable aspect of a high-quality web application is the use of client-side
JavaScript to extend the application directly into the client web browser. Incorporating JavaScript behaviors into the user interface allows for interactions that
simply aren’t possible using the standard approaches involving links and form
submissions. The interface can react to the user without the delay of a new
request to the server.
The downside to making use of client-side scripting is that the skills necessary to write functional, cross-browser JavaScript are extremely specialized, bordering on arcane, and not something every developer has accumulated. Using
Tapestry unavoidably complicates this further, because Tapestry pages can be so
very dynamic.
You’ve already seen examples of client-side logic in Tapestry. In chapter 4,
you saw the DatePicker component, which uses client-side JavaScript to create a
pop-up calendar. In chapter 5, you saw that the input validation framework
could perform client-side validations before forms were submitted, in addition to
performing server-side validations of the submitted form data.
You’ve also seen the Palette component, in chapter 6, which uses a significant
amount of client-side scripting to provide a complicated user interface. In all of
these cases, the rendering of HTML also involved the production of customized
JavaScript to support the HTML elements. For the validation framework, the
JavaScript took the form of onsubmit form event handlers that validated user
input and displayed alerts if the input was not acceptable. The client-side scripting for the Palette component includes much logic for enabling and disabling
buttons and moving selections between the two columns. This is much more
340
CHAPTER 8
Advanced techniques
than simply including a pre-canned JavaScript file; in each of these cases, it is
necessary to create the client-side JavaScript dynamically, using a template, and
adapt it to the specific situation it is being used for.
In the case of ValidField components, the exact checks that will occur vary
from one field to the next. Beyond general format checks, there is checking for
minimum values, maximum values, and so forth. Each of these checks is configured on a case-by-case basis. In addition, the JavaScript generated for a ValidField component must adapt to the element name for the component (which is
assigned by the enclosing Form) and to the name of the enclosing Form itself
(which is assigned by Tapestry).
In a traditional servlet application, the development process often consists of
creating the script and then requiring that the names of the elements on the
page (the form, the input field) match. This is certainly simpler—if the usage of
the script is well understood and well documented. All too frequently, that isn’t
the case; JavaScript support is often provided by the HTML designers, who are,
naturally enough, the least disciplined team members when it comes to creating
code. In addition, it is too easy to create a script that specifically allows one field
in one form to have special behavior, and that is not able to handle multiple
fields in multiple forms.
As elsewhere, Tapestry tips this equation on its head. Tapestry uses a special
template, a script specification, to create customized JavaScript on demand.
Even in the simplest case, the script specification must adapt to the various IDs
assigned by the framework. This means that a Tapestry script specification easily
adapts to reuse within the same page. The downside is that instead of writing a
block of JavaScript, you must write a template from which the JavaScript can be
dynamically produced at runtime.
The other great benefit of dynamic JavaScript generation is that the creation
of the JavaScript is compartmentalized inside specific components. Once again,
the component may be dropped into a page and will function properly, clientside JavaScript and all, without any special handling in the page. By using JavaScript-enabled components, you gain all the advantages of a client-side JavaScript
without any of the costs!
Creating such a component has a few prerequisites:
■
You must know exactly what the component should do.
■
You must understand how to accomplish that using JavaScript, which
includes understanding the client-side Document Object Model (DOM)
and the programming differences between various browsers.
■
You must understand how Tapestry assigns names and IDs.
■
You must understand how to invoke the Tapestry scripting subsystem.
341
As usual, we’ll start with a reasonable example.
8.2.1 Defining the CreditCardField component
Many web applications have some point where they ask for a credit card number.
All too often, you are advised on how to enter that number: as all digits, no
spaces or dashes. Forcing the user to bend to the needs of the programmer is
never a good sign; instead, the interface should resemble figure 8.5. Here, the
input is four individual text fields. As you fill one field, the cursor automatically
tabs to the next field. When you submit the form, the input is checked for validity
(that each of the four fields has exactly four numeric digits).
Let’s develop this component in three steps:
Figure 8.5 The CreditCardField component renders as four separate text fields. It includes clientside scripting to automatically tab to the next field as each field is filled. When the form is submitted,
client-side scripting checks that each field consists of exactly four numeric digits.
342
CHAPTER 8
Advanced techniques
■
We’ll create a static HTML mockup, in order to figure out the general form
of the client-side JavaScript.
■
We’ll create a Tapestry script specification, a special-purpose template for
creating JavaScript dynamically.
■
Finally, we’ll create a component that makes use of the script specification
when it renders and knows how to obtain the submitted value when the
enclosing form is submitted.
Let’s start with the HTML mockup, shown in listing 8.7.
Listing 8.7 HTML mockup for the CreditCardField component
<html>
<body>
<script>
function onkeyup_cc_field(field, next)
{
var keycode = window.event.keyCode;
if (keycode >= 48 && keycode <= 57 && field.value.length == 4)
{
next.focus();
next.select();
}
}
c
function validate_cc_field(field)
{
if (! field.value.match(/^\d{4}$/))
{
field.focus();
field.select();
window.alert(
"Credit card numbers consist of four groups of " +
"four numbers.");
return false;
}
return true;
}
d
function onkeyup_cc$g0 ()
{
onkeyup_cc_field(document.f.cc$g0, document.f.cc$g1);
}
b
343
{
}
{
}
e
function validate_cc()
{
return validate_cc_field(document.f.cc$g0) &&
validate_cc_field(document.f.cc$g1) &&
validate_cc_field(document.f.cc$g2) &&
validate_cc_field(document.f.cc$g3);
}
f
window.onload = function ()
{
document.f.cc$g0.onkeyup = onkeyup_cc$g0;
document.f.onsubmit = validate_cc;
}
</script>
<form name="f">
Enter credit card number:
<input type="text" name="cc$g0"
value=""
value=""
value=""
value=""
size="4"
size="4"
size="4"
size="4"
maxlength="4"/> maxlength="4"/> maxlength="4"/> maxlength="4"/>
<br/>
</form>
</body>
<html>
b
c
d
When a number is entered and the field is full, the cursor moves to the next field.
This JavaScript function is invoked when the form submits to validate that the
field contains exactly four digits.
Each of the first three fields gets its own onkeyup event handler.
344
CHAPTER 8
Advanced techniques
e
f
When the form submits, it invokes the validate_cc() function to validate that
each of the four fields contains exactly four digits.
When the HTML page finishes loading into the web browser, this anonymous
function registers event handlers for the form and fields.
This listing may look somewhat odd, since it’s very mechanical in structure. This
is to reflect the constraints on generating the JavaScript that occur in a running
Tapestry application. The path of least resistance is to attach JavaScript event
handlers to objects from within the window.onload event handler, rather than as
javascript: attribute values on the elements themselves. As you’ll see, the script
for a component is often generated after the component itself renders, when it is
too late to provide values for attributes.
In addition, the names assigned to the four text fields reflects some experience concerning how dynamic names may be generated within Tapestry. Here,
the cc portion of the name will probably be a component ID. The suffix (g0, g1,
etc.) distinguishes the four groups of digits that compose a credit card number
(this component is focused on 16-digit credit card numbers, such as used by
Master Card and Visa; other credit cards use a different format). Using the dollar
sign as a separator is a safe bet: the dollar sign is not allowed in a user-specified
component ID, so there’s no possibility of a naming conflict.
The operation of the mockup script is simple enough. The onkeyup event
handlers are triggered as each key is released within the first three fields. If the
key entered was a number and that number filled the field, the cursor moves to
the next field. When the form submits, the onsubmit event handler checks each
of the four fields against a regular expression to ensure that they contain exactly
four numeric digits.
8.2.2 Working with the Body component
As mentioned in chapter 4, when a page uses dynamic JavaScript, you must use a
Body component to generate the HTML <body> element. As this chapter demonstrates, it’s not always possible to know ahead of time if a component will be
using the JavaScript features provided by the Body component, so you should
always use a Body component, just to be safe.
A principal function of the Body component is to support and organize the
generation of dynamic JavaScript throughout the page. All the JavaScript created by all components throughout the page is organized into a single HTML
<script> block, which is placed just inside the HTML <body> element, as shown
in figure 8.6.
345
Figure 8.6 The Body component organizes all the JavaScript for the page into a
single block, just inside the <body> tag. Any component enclosed by the Body
may contribute to the script body or script initialization.
Notice that the scripting block goes first even though what goes into the scripting block is determined by rendering the body of the Body component. The
Body component buffers the output of the components it encloses. As each component renders, it has a chance to ask the Body component to contribute some
text to the script body or script initialization.
Once rendering is complete, the Body component can render out the <script>
block, then the buffered output from its body and then, finally, the end tag. This
approach—buffering the output of enclosed components to affect how the enclosing component renders—is a powerful technique used throughout Tapestry.
8.2.3 Creating the Tapestry script specification
Creating the Tapestry script specification and the component would normally
occur in parallel; they are both dependent on each other. For this example, we’ll
show you how to convert the static JavaScript from the mockup into a dynamic
346
CHAPTER 8
Advanced techniques
script specification. At runtime, the XML script specification will be parsed into
an executable script.
When a script is executed, it accomplishes three things:
■
Input symbols are combined and manipulated to create new symbols. For
example, the component’s element ID may be used as the basis of a JavaScript event handler function name.
■
A contribution to the page’s script body block is made. This is typically
JavaScript event handler functions.
■
A contribution to the page’s script initialization block is made. This is typically setup code that connects client-side DOM objects to the JavaScript
event handler functions.
Listing 8.8 shows the script specification used by the CreditCardField component. Large sections of it are similar to the static HTML from listing 8.7. The
remainder involves customizing each usage of the script to the particulars of the
page where it is used and the component that uses it.
Listing 8.8
CreditCardField.script: script specification used by the component
<!DOCTYPE script PUBLIC
"-//Apache Software Foundation//Tapestry Script Specification
3.0//EN"
"http://jakarta.apache.org/tapestry/dtd/Script_3_0.dtd">
➥
<script>
<input-symbol
key="creditCardField"
class="examples.cc.CreditCardField" required="yes"/>
<input-symbol key="formatMessage"
class="java.lang.String"
required="yes"/>
c
<set key="name" expression="creditCardField.name"/>
<let key="baseName">
document.${creditCardField.form.name}.${name}
</let>
<let key="field0">
${baseName}$g0
</let>
b
d
e
f
347
<let key="field1">
${baseName}$g1
</let>
<let key="field2">
${baseName}$g2
</let>
<let key="field3">
${baseName}$g3
</let>
<let key="formValidateFunction" unique="yes">
validate_${name}
</let>
g
<body>
<unique>
<![CDATA[
{
{
next.focus();
next.select();
}
}
{
{
field.focus();
field.select();
window.alert("${formatMessage}");
return false;
}
return true;
}
]]>
</unique>
<![CDATA[
function ${formValidateFunction}()
{
return validate_cc_field(${field0}) &&
validate_cc_field(${field1}) &&
h
348
CHAPTER 8
Advanced techniques
validate_cc_field(${field3});
}
]]>
</body>
<initialization>
${field0}.onkeyup = function()
{
onkeyup_cc_field(${field0}, ${field1});
}
{
}
{
}
</initialization>
</script>
b
c
d
e
f
g
h
The <input-symbol> element is equivalent to a component specification <parameter> element. It declares a particular type of symbol to be passed into the
script, with a type. The creditCardField symbol must be an instance of CreditCardField, the Java class for the CreditCardField component.
The formatMessage symbol is required and must be a string.
The <set> element is used to create a new symbol using an OGNL expression.
The <let> element creates a new symbol from a string. Leading and trailing
white space is removed. The ${…} values are interpreted as OGNL expressions;
this is how other symbols are referenced.
The component and the script must agree on what each of the four fields is
named. The name is based on the field’s element ID with $g0, $g1, $g2, or
$g3 appended.
Tapestry can ensure that the new symbol value is unique within the page, which
may require appending a suffix to the value.
The contents of a <unique> element will be rendered only once per page render,
even if the script specification is used multiple times on the same page. The XML
<![CDATA[…]]> construct is used to work around the & characters, which are not
normally valid XML.
349
When the script executes, it takes input symbols (primarily, the CreditCardField
component for which the script is being generated), transforms the input symbols into new symbols (such as the field and function names), and generates JavaScript for inclusion in the page. In some cases, the created symbols are useful to,
and needed by, the component.
Appendix D includes a complete definition of the script specification DTD.
Defining the input symbols
The script begins by defining two input symbols, which must be provided by the
CreditCardComponent when it executes the script. The first symbol, creditCardField, is the Java object for the component itself. The second, formatMessage,
is a message that will be displayed to users if they attempt to enter a credit card
number using an invalid format.
Defining input symbols is optional; it simply allows Tapestry to check that
required symbols are provided. If a class is specified, then the symbol must be
that type, or assignable to that type (it may be the fully qualified name of a class
or interface). Again, providing this information to Tapestry allows the framework
to perform better error checking.
These symbols, especially creditCardField, become the basis for all the other
symbols created while the script is executing.
Creating new symbols
New symbols are created using the <set> and <let> elements. The <set> element
is used to evaluate an OGNL expression and assign its value as a new symbol. It is
a close cousin to the <binding> element in a page or component specification. In
the CreditCardField script, the name symbol is created by retrieving the name
property of the creditCardField symbol:
<set key="name" expression="creditCardField.name"/>
If the OGNL expression is long or complicated, then the expression attribute can be
omitted, and the OGNL expression can be placed in the body of the <set> element.
The <let> element creates a new symbol, always a string, from its body, but its
body can contain both text and specially delimited OGNL expressions.1 In the
script, the baseName symbol is created using a <let> element:
1
The <let> element may also contain several other script elements: <foreach> for looping, <if> and <ifnot> for conditionals, and the <unique> element we’ll cover shortly—but using any of these is quite rare.
350
CHAPTER 8
Advanced techniques
<let key="baseName">
document.${creditCardField.form.name}.${name}
</let>
This baseName symbol is a JavaScript DOM reference to a field. The ${…} sequences
indicate OGNL expressions. The first traverses from the creditCardField to the
Form component that encloses it and obtains the name assigned to the Form.
The second expression references the previously defined name symbol. Ultimately, this symbol is used to build the complete names of the four text fields
that form the component. Table 8.1 identifies the remaining symbols created by
the script (the Example Value column assumes the Form’s name is Form0 and
the CreditCardField’s name is ccField, though of course the actual names will
vary from page to page).
Table 8.1
Symbols created by the script
Symbol
Description
Example Value
name
Name assigned to the CreditCardField by the Form
ccField
baseName
Base value for building clientside DOM references to the four
text fields
document.Form0.ccField
field0
First field’s DOM reference
document.Form0.ccField$g0
field1
Second field’s DOM reference
field2
Third field’s DOM reference
field3
Fourth field’s DOM reference
formValidateFunction
Name of the JavaScript event
handler for the containing
Form’s onsubmit event
validate_ccField
A few notes:
■
The four text fields are given names based on the component. The dollar
sign is a good choice for a separator here, since it is not allowed as a part of
a component ID—this ensures that there will be no name conflicts, regardless of which component IDs the developer selects.
■
The naming of the fields must coordinate with code in the component that
writes the HTML for the four text fields and reads the query parameters
when the form is submitted.
■
351
The formValidateFunction is an output symbol, the name of a JavaScript
function. Shortly, you’ll see how to hook it into the enclosing form’s onsubmit
event handler.
The <let> element that defines the formValidateFunction symbol demonstrates
another useful feature:
<let key="formValidateFunction" unique="yes">
validate_${name}
</let>
It is very important that this function name be unique within the page; if the
CreditCardField component is used multiple times, then each usage of the component must have its own validate function. Simply incorporating the component’s name into the name of the function is very nearly enough to ensure
uniqueness because the name itself is uniquely generated by the form.
To absolutely ensure that the value for the formValidateFunction symbol is
unique, we’ve used the unique attribute of the <let> element. Tapestry ensures
that the final symbol value is unique, tacking on a suffix to ensure uniqueness if
necessary. This functionality is provided by the Body component that ultimately
encloses the CreditCardField; therefore, it doesn’t matter which script specification is executed—the uniqueness filter is applied universally.
NOTE
The use of the unique attribute here is overkill. It is all but impossible
for two different CreditCardFields to share the same element ID, because element IDs are related to component IDs, which are always
unique within a container (the page or component). Uniqueness issues
tend to crop up when a complex component is assembled from simpler
components, and the complex component is used multiple times—the
Palette component is an example of a complex component that requires
uniqueness support to ensure that you can use it faultlessly in all circumstances. Tapestry provides you with the tools to defuse this potential issue painlessly.
TIP
Use reasonable names, and incorporate the component ID into the function name. You may have to debug the rendered HTML page, and it will
be easier to do so if the function names reflect the component they are
related to.
352
CHAPTER 8
Advanced techniques
Contributing to the body
The script specification <body> element is where the real work begins. This is
where the JavaScript event handlers are created. These event handlers will provide two examples of client-side behavior:
■
■
When the user fills one of the component’s text fields, the cursor will automatically tab to the next text field.
When the user submits the form, the text fields will be validated to contain
exactly four numeric digits.
The CreditCardField script includes some common utility functions: onkeyup_
cc_field() and validate_cc_field(). These two functions are used by all four
text fields created by the component—but potentially, there will be other CreditCardField components on the same page, and they all should share the same
utility functions. These utility functions should be contributed into the page’s
<script> block only once.
This situation is similar to the previously described issue concerning unique
symbols, and it has a similar solution. The portion of the <body> element that
should only occur once is itself enclosed in a <unique> element:
<unique>
<![CDATA[
{
{
next.focus();
next.select();
}
}
{
{
field.focus();
field.select();
window.alert("${formatMessage}");
return false;
}
return true;
}
]]>
</unique>
353
When the <unique> element is encountered, it checks to see if the block it
encloses has been contributed yet; if not, the contents of the <unique> element
are contributed to the page’s <script> block, and the fact that the block has been
contributed is recorded. As with the unique symbol generation, this tracking is a
responsibility of the Body component, and it applies only to the current request
cycle. Whatever is enclosed by the <unique> element will appear just once inside
the page’s <script> block.
The remainder of the <body> element’s content is not enclosed by the
<unique> tag and is contributed to the page’s <script> block every time the script
specification is executed. This is where the form’s validate function is created:
<![CDATA[
function ${formValidateFunction}()
{
return validate_cc_field(${field0}) &&
validate_cc_field(${field3});
}
]]>
This snippet uses a CDATA block to work around the invalid XML characters: the
&& (logical and) operators that are valid JavaScript but not acceptable in an XML
document. As before, we use the ${…} syntax to plug in the symbols created earlier in the script specification.
In the final rendered page, with the expansion of symbols, this block will look
something like this:
function validate_ccField()
{
return validate_cc_field(document.Form0.ccField$g0) &&
validate_cc_field(document.Form0.ccField$g1) &&
validate_cc_field(document.Form0.ccField$g2) &&
validate_cc_field(document.Form0.ccField$g3);
}
Of course, both the name of the form (Form0) and the name of the component
(ccField) will vary.
Wiring event handler methods to client-side objects
The last step is to connect event handler functions to the objects. This poses a
minor dilemma: The HTML <script> element containing all the JavaScript for the
page is placed just inside the HTML <body> element. The HTML that ultimately
354
CHAPTER 8
Advanced techniques
defines the objects comes later. Referencing those objects using the DOM before
they exist will cause runtime errors inside the client web browser, and your clientside logic will break.
The solution to this is to wait until the page is fully loaded into the client web
browser before accessing the objects and connecting up event handlers. That is
the function of the <initialization> element in the script specification. The
JavaScript contributed inside the <initialization> element is executed within
the window.onload event handler, which is invoked only after the page is fully
loaded. The script specification includes an <initialization> block to “wire up”
the handlers by setting the event handlers (such as onkeyup) for the fields:
<initialization>
{
}
{
}
{
}
</initialization>
This wires up the first three (of the four) fields, creating anonymous JavaScript
functions that invoke the onkeyup_cc_field() function. This function checks to
see if the pressed key was a number and if the field is now full, and tabs to the
next form field if that’s the case. Unfortunately, we can’t tab out of the fourth
text field, since the CreditCardField doesn’t know what the next field is (it will
not have even rendered yet at the time the CreditCardField component renders
and executes the script specification).
That still leaves the form’s submit function, which checks that each of the four
fields contains exactly four numeric digits. In theory, we could hook that in here
as well:
document.${creditCardField.form.name}.onsubmit =
${formValidateFunction};
That approach would work—as long as the CreditCardField component was
the only component within the Form that needed to hook into the Form’s
355
onsubmit event handler. There’s no way the CreditCardField can know this,
however; certainly there could be ValidField components with client-side validation enabled, as well as other custom components, which may render before
or after the CreditCardField.
Instead, this problem is solved at the code level as an interaction, in Java
code, between the CreditCardField component and the Form component that
encloses it. The Form component is prepared to handle the case where multiple components need to hook into the onsubmit event handler. The details of
this are covered shortly, along with the code for the CreditCardComponent
Java class.
8.2.4 Creating the CreditCardField specification
The CreditCardField component is a form element component; it is intended
for use as part of a form. As such, its Java class has some basic responsibilities:
■
The class must implement the interface IFormComponent.
■
The class must implement a form property of type IForm.
■
The class must implement a name property of type string.
The easiest way to accomplish all this is to have the CreditCardField class inherit
from the AbstractFormComponent base class, and to specify the two properties
(form and name) in the component specification. Inheriting from AbstractFormComponent adds another requirement: implementing a disabled parameter of
type boolean.
In addition, the whole point of the CreditCardField component is to allow
the user to enter a credit card number and update a domain object property.
Although the user interface is in the form of four text fields, each accepting
four digits, this is not the best way to represent the credit card number once
it’s submitted in the form. A single 16-digit string is much more natural. The
CreditCardField defines this, as the parameter cardNumber. This is another
example of the Model-View-Controller pattern: The Model (the domain
object property) represents the credit card number as a single string; the View
(the HTML rendered by the component) uses four individual fields. The Controller, the CreditCardField class, is responsible for translating between the
two representations.
Listing 8.9 shows how the CreditCardField class specifies its two parameters
(cardNumber and disabled) and two properties (form and name).
356
CHAPTER 8
Advanced techniques
Listing 8.9 CreditCardField.jwc: specification for the CreditCardField component
class="examples.cc.CreditCardField"
allow-body="no" allow-informal-parameters="no">
<parameter
name="cardNumber"
direction="form"
required="yes"/>
<parameter name="disabled" type="boolean" direction="in">
<description>
If true, then all four text fields will be disabled.
</description>
</parameter>
<property-specification
name="form"
type="org.apache.tapestry.IForm"/>
name="name"
The cardNumber parameter uses the value form for its direction attribute, which
makes sense for this purpose, a form element component. The property bound
to the cardNumber parameter will be updated (from the cardNumber property)
after the CreditCardComponent rewinds when the enclosing form is submitted.
The disabled parameter still uses in as its direction attribute; the disabled
property will be set from the parameter before the component renders (even
during a form rewind).
8.2.5 Creating the CreditCardField component
The CreditCardField Java class is shown in listing 8.10. CreditCardField extends
from AbstractFormComponent, a base class used by many different form element
components (such as Checkbox and TextField). AbstractFormComponent defines
two abstract properties, form and name, which we must implement (by specifying
357
those properties in the component specification). In addition, CreditCardField
is required to implement a disabled parameter.
Listing 8.10 CreditCardField.java: Java class for the CreditCardField component
package examples.cc;
import java.util.HashMap;
import java.util.Map;
import
import
import
import
import
import
import
import
import
import
import
import
import
import
import
import
import
import
org.apache.oro.text.regex.MalformedPatternException;
org.apache.oro.text.regex.Pattern;
org.apache.oro.text.regex.Perl5Compiler;
org.apache.oro.text.regex.Perl5Matcher;
org.apache.tapestry.IForm;
org.apache.tapestry.IResourceLocation;
org.apache.tapestry.IScript;
org.apache.tapestry.engine.IScriptSource;
org.apache.tapestry.form.AbstractFormComponent;
org.apache.tapestry.form.FormEventType;
org.apache.tapestry.html.Body;
org.apache.tapestry.request.RequestContext;
org.apache.tapestry.valid.ValidationConstraint;
org.apache.tapestry.valid.ValidatorException;
public abstract class CreditCardField
extends AbstractFormComponent
{
private IScript _script;
private Pattern _compiledPattern;
private Perl5Matcher _matcher;
public abstract void setCardNumber(String cardNumber);
public abstract String getCardNumber();
{
IForm form = getForm(cycle);
b
String name = form.getElementId(this);
boolean disabled = isDisabled();
IValidationDelegate delegate = form.getDelegate();
if (form.isRewinding())
{
d
c
358
CHAPTER 8
Advanced techniques
if (!disabled)
updateCardNumberFromRequest(name, form, delegate, cycle);
return;
}
return;
String cardNumber[] = extractCardNumber(delegate);
delegate.writePrefix(writer, cycle, this, null);
e
f
for (int i = 0; i < 4; i++)
{
if (i > 0)
writer.print(" - ");
String fieldName = name + "$g" + i;
g
writer.beginEmpty("input");
writer.attribute("type", "text");
writer.attribute("name", fieldName);
writer.attribute("size", 4);
writer.attribute("maxlength", 4);
writer.attribute("value", cardNumber[i]);
if (disabled)
writer.attribute("disabled", "disabled");
delegate.writeAttributes(writer, cycle, this, null);
}
delegate.writeSuffix(writer, cycle, this, null);
if (!disabled)
{
if (_script == null)
{
IScriptSource source = cycle.getEngine().getScriptSource();
IResourceLocation specLocation =
getSpecification().getLocation().getResourceLocation();
IResourceLocation scriptLocation =
specLocation.getRelativeLocation(
"CreditCardField.script");
_script = source.getScript(scriptLocation);
}
Body body = Body.get(cycle);
h
Map symbols = new HashMap();
symbols.put("creditCardField", this);
symbols.put("formatMessage",
getMessage("card-number-format"));
_script.execute(cycle, body, symbols);
i
j
String formValidateFunction =
(String) symbols.get("formValidateFunction");
form.addEventHandler(FormEventType.SUBMIT,
formValidateFunction);
1)
}
}
private void updateCardNumberFromRequest(
String name,
IForm form,
{
RequestContext context = cycle.getRequestContext();
for (int i = 0; i < 4; i++)
{
String value = context.getParameter(name + "$g" + i);
if (value != null)
{
if (value.length() > 4)
value = value.substring(0, 4);
buffer.append(value);
}
while (buffer.length() < 4 * (i + 1))
buffer.append(' ');
}
String cardNumber = buffer.toString();
delegate.recordFieldInputValue(cardNumber);
try
{
validate(cardNumber);
setCardNumber(cardNumber);
}
catch (ValidatorException ex)
1!
1@
359
360
CHAPTER 8
Advanced techniques
{
delegate.record(ex);
}
}
private String[] extractCardNumber(IValidationDelegate delegate)
{
String cardNumber =
delegate.getFieldInputValue();
1#
if (cardNumber == null)
cardNumber = getCardNumber();
return new String[] { "", "", "", "" };
String[] result = new String[4];
StringBuffer buffer = new StringBuffer(cardNumber);
while (buffer.length() < 16)
buffer.append(' ');
for (int i = 0; i < 4; i++)
result[i] = buffer.substring(4 * i, 4 * (i + 1));
return result;
}
private void validate(String cardNumber)
throws ValidatorException
{
if (_compiledPattern == null)
{
Perl5Compiler compiler = new Perl5Compiler();
try
{
_compiledPattern = compiler.compile("^\\d{16}$");
}
catch (MalformedPatternException ex)
{
}
}
if (_matcher == null)
_matcher = new Perl5Matcher();
if (_matcher.matches(cardNumber, _compiledPattern))
return;
361
String formatted =
cardNumber.substring(0, 4)
+ "-"
+ cardNumber.substring(4, 8)
+ "-"
+ cardNumber.substring(8, 12)
+ "-"
+ cardNumber.substring(12);
throw new ValidatorException(
format("invalid-card-number", formatted),
ValidationConstraint.NUMBER_FORMAT);
}
}
b
c
d
e
f
g
h
i
j
1)
The inherited getForm() method locates the enclosing Form component (represented as an instance of the IForm interface). From the Form, we obtain the name
used for this form control component.
This component integrates into the validation subsystem (from chapter 5) and
needs access to the validation delegate, which is provided by the Form component.
When the Form is rewinding, we don’t render HTML. Instead, we invoke this
method to assemble the 16-digit credit card number from the four query parameters (one for each text field in the rendered form), as well as perform validations.
When rendering, the 16-digit credit card number must be split into four groups
of 4 digits.
As with a ValidField component, the validation delegate is allowed to render
before and after the text fields are rendered. This allows the validation delegate
to write HTML to support any application-specific look and feel.
The names for the individual text fields computed here must match against the
JavaScript generated by the script specification.
The script source is used to obtain an executable script. The location of the script is
computed relative to the component specification. Because of these extra steps
required to obtain the executable script instance, the instance is cached for later use.
Input symbols are provided to the executable script as a Map.
The script is executed. It will modify the symbols Map as it executes. It will also
communicate with the Body component to contribute JavaScript to the page’s
script block.
The formValidateFunction symbol is created by the script to store the name of a
client-side JavaScript function that must be hooked into the form to validate the
fields. The addEventHandler() method is used to inform the enclosing Form
component about the function.
362
CHAPTER 8
Advanced techniques
1!
1@
1#
The validation delegate can store only a single input string, so the 16-digit credit
card number is recorded. If the client has disabled JavaScript, then this value
may not be a valid credit card number and will be used when the form renders to
present errors.
As with a ValidField, the input is validated before the final value (a valid 16-digit
credit card number) is assigned to the cardNumber parameter.
If the validation delegate has a recorded input value for this component, then
use that value (it represents a value provided by the user). Otherwise, obtain the
value from the cardNumber parameter.
Even though CreditCardField is not a ValidField, it expects the form to have a
delegate and will make use of it to report any input errors. This can happen only
if the client web browser fails to do its part, probably because the user has explicitly disabled JavaScript. In any case, if invalid data does make it to the server, the
component will perform its own validations, but will interact with the validation
delegate to visually show that the component (all four fields) is in error. Figure 8.7
shows an example of this.
Figure 8.7 The component does its own validations, in case JavaScript is not enabled in the client.
The component interacts with the input validation subsystem to adjust the visual display of the
fields to reflect that the component is in error.
363
The CreditCardComponent has a number of responsibilities: It must render the
HTML for the four credit card fields as well as the JavaScript that supports the
four fields, and it must handle form submissions for the fields. It also translates
between the client-side data format (four fields of 4 digits) and the server-side
data format (a single 16-digit string).
Rendering the component
Because this is a form element component, the first step in the renderComponent()
method is to find the enclosing Form component and have the Form provide a
name for this element. The AbstractFormComponent base class provides most of
this functionality:
IForm form = getForm(cycle);
String name = form.getElementId(this);
boolean disabled = isDisabled();
IValidationDelegate delegate = form.getDelegate();
If the form is rewinding, we extract the credit card number from the query
parameters, validate the number, and return:
if (form.isRewinding())
{
if (!disabled)
updateCardNumberFromRequest(name, form, delegate, cycle);
return;
}
Alternately, if the page as a whole is rewinding but not specifically this form, we
can skip the rest. This scenario occurs only when the action service is used:
return;
Next we render the four fields and the dashes that separate them. The first step
is to extract the 16-digit credit card number and break it apart into four strings,
one for each 4-digit group. We can then render each group as its own text field:
String cardNumber[] = extractCardNumber(delegate);
delegate.writePrefix(writer, cycle, this, null);
for (int i = 0; i < 4; i++)
{
if (i > 0)
writer.print(" - ");
364
CHAPTER 8
Advanced techniques
String fieldName = name + "$g" + i;
writer.beginEmpty("input");
writer.attribute("type", "text");
writer.attribute("name", fieldName);
writer.attribute("size", 4);
writer.attribute("maxlength", 4);
writer.attribute("value", cardNumber[i]);
if (disabled)
writer.attribute("disabled", "disabled");
delegate.writeAttributes(writer, cycle, this, null);
}
delegate.writeSuffix(writer, cycle, this, null);
In addition, we invoke methods on the validation delegate to decorate our fields
if they are in error.
Converting the server-side value
The credit card number stored as a server-side property is a 16-digit string, but
when rendering the four text fields, four smaller (4-digit) strings are needed.
This is handled in the extractCardNumber() method.
In addition, another interaction with the input validation subsystem occurs
here. When a form is submitted, the validation delegate records the input provided by the user, before the values are converted and validated. If the form is
redisplayed (due to an error), the validation delegate is the source of the unconverted values. The validation delegate’s getFieldInputValue() method is used to
gain access to the exact input provided by the user so that the invalid input can
be sent back to the user for correction:
private String[] extractCardNumber(IValidationDelegate delegate)
{
String cardNumber = delegate.getFieldInputValue();
cardNumber = getCardNumber();
return new String[] { "", "", "", "" };
String[] result = new String[4];
StringBuffer buffer = new StringBuffer(cardNumber);
while (buffer.length() < 16)
buffer.append(' ');
365
for (int i = 0; i < 4; i++)
result[i] = buffer.substring(4 * i, 4 * (i + 1));
return result;
}
So far, we’ve covered somewhat familiar ground; CreditCardComponent is simply a complicated component. What makes it interesting is what comes next:
adding the client-side JavaScript.
Obtaining the script
The final portion of the renderComponent() method involves obtaining and executing the script specification. The script specification is stored in the WEB-INF
folder alongside the component specification and is named CreditCardField.script. The script specification must be parsed and converted into an executable script: an instance of the IScript interface.
Scripts are obtained from the script source, a Tapestry subsystem accessible
via the engine. To get a script, we must know the location of its script specification. We are able to build a location relative to the component’s specification:
if (_script == null)
{
IScriptSource source = cycle.getEngine().getScriptSource();
IResourceLocation specLocation =
getSpecification().getLocation().getResourceLocation();
IResourceLocation scriptLocation =
specLocation.getRelativeLocation("CreditCardField.script");
_script = source.getScript(scriptLocation);
}
It’s likely that the page containing the CreditCardComponent will be used many
times, so we cache the IScript instance in the _script instance variable. The
script source also caches parsed script specifications (once it parses a script specification, it doesn’t parse it a second time), so all we’re saving is the runtime cost
of constructing the script specification’s location.
Executing the script
To execute the script, we need to get the Body component. The Body class
includes a static method for accomplishing this:
Next we must create the input symbols that are passed to the script when it executes:
366
CHAPTER 8
Advanced techniques
Map symbols = new HashMap();
symbols.put("creditCardField", this);
symbols.put("formatMessage", getMessage("card-number-format"));
With everything in place, we can now execute the script, which will communicate
with the Body component:
_script.execute(cycle, body, symbols);
The final step is to connect the form’s validate function to the Form component.
The script will have modified the symbols Map passed to it. The Map now includes
all the symbols created using the <let> and <set> elements, including the formValidateFunction symbol. It is easy enough to extract the value for that symbol,
which is the name of the JavaScript function:
String formValidateFunction =
(String) symbols.get("formValidateFunction");
form.addEventHandler(FormEventType.SUBMIT, formValidateFunction);
The Form component includes an addEventHandler() method that allows the
components it encloses to hook into its onsubmit event handler. Any number of
components may provide the names of JavaScript event handler functions to
execute; the Form component will execute each in turn.
Handling form submissions
When the form containing the CreditCardField component is submitted, the CreditCardField is responsible for reading the query parameters submitted as part of the
form, assembling the 16-digit card number from the four text fields, validating the
card number, and updating the creditCard parameter property with the final value.
Most of this logic is present in the updateCardNumberFromRequest() method. It
makes use of the RequestContext object to access the query parameters posted by
the form:2
private void updateCardNumberFromRequest(
String name,
IForm form,
{
2
Although you can also get access to the HttpServletRequest to read query parameters, it is best to
use the RequestContext object. Because of the way the Servlet API handles file uploads (as discussed
in chapter 4), when a form contains an Upload component, the HttpServletRequest object will not
have access to the query parameters, but the RequestContext object will.
367
RequestContext context = cycle.getRequestContext();
for (int i = 0; i < 4; i++)
{
String value = context.getParameter(name + "$g" + i);
if (value != null)
{
if (value.length() > 4)
value = value.substring(0, 4);
buffer.append(value);
}
while (buffer.length() < 4 * (i + 1))
buffer.append(' ');
}
String cardNumber = buffer.toString();
delegate.recordFieldInputValue(cardNumber);
try
{
validate(cardNumber);
setCardNumber(cardNumber);
}
catch (ValidatorException ex)
{
delegate.record(ex);
}
}
The method has extra checks on the length and format of the credit card number that duplicate the client-side checks. This is still appropriate, since there is
no way to ensure that the client-side validations took place (the client web
browser may have JavaScript disabled, or may not even be a web browser).
A key part of the validation subsystem is the IValidationDelegate’s recordFieldInputValue() method. This is used to record the value of a submitted ValidField before it is converted to a target type and validated. This explains how,
when a form is redisplayed with errors, the input provided by the user can be
redisplayed as entered; the input “lives” inside the validation delegate until
needed to redisplay the form.
Since the CreditCardField should mimic the behavior of a ValidField component, it too must inform the validation delegate about the user’s input.
368
CHAPTER 8
Advanced techniques
Before updating the cardNumber parameter property, the complete credit card
number is validated, using the validate() method. This method checks that the
number consists of 16 digits; if it doesn’t, the method formats an error message
and throws a ValidatorException. The updateCardNumberFromRequest() method
catches that exception and passes it to the Form’s validation delegate.
8.2.6 Using the component
So far, we’ve shown how to create a complex component, a component that renders a complex mix of HTML and JavaScript. Despite this, the component fits
into the overall page’s HTML template very simply:
<tr>
<th>
<span jwcid="@FieldLabel"
displayName="Credit Card Number"
field="ognl:components.inputNumber">
Credit Card Number
</span>
</th>
<td>
<input type="text"
jwcid="inputNumber@CreditCardField"
cardNumber="ognl:cardNumber"
size="50"/>
</td>
</tr>
As you can see, using the CreditCardField component is no more complex than
using a ValidField component. We’re using a FieldLabel with the CreditCardField, just as we did in chapter 5. When previewing the page, the component will
display as a single, 50-character wide text field (even though in the live application, it renders as four small text fields).
Adding client-side JavaScript support to a component is one of the more
complex tasks you will face as a Tapestry developer—but by keeping a clear head
about which code (Java or JavaScript) is executed where (on the server or in the
client web browser) and when (at page render, within the client, or on form submission), you’ll ensure that the entire process is manageable. Tapestry includes
the necessary subsystems and hooks to provide you with the flexibility you need to
create powerful, reusable components.
Up to this point, we’ve discussed integrating components with Tapestry
engine services and integrating components with dynamically generated JavaScript. Another common task is to integrate a Tapestry application with a
non-Tapestry application.
369
8.3 Integrating with JavaServer Pages
There will be times when you must make a Tapestry application work in concert
with a standard servlets-and-JSP application. Most likely, you have an existing
application, and you want to transition to a Tapestry implementation but can’t
do it all in one fell swoop. What you need is a hybrid application, partly written
using Tapestry and partly written using servlets and JSPs.
In some cases, JSPs will have links that connect to Tapestry pages, or forms
that submit to Tapestry pages. On the flip side, you may have listener methods
that use a JSP (or even some other servlet) to render a response. Tapestry supports all of these scenarios.
We’ll explore all of these possibilities using the sample application; you can
open a web browser to http://localhost:8080/examples/app and select the JSP link
under the chapter 8 heading.
8.3.1 Redirecting to a JSP
On the Tapestry side, you may need to use a JSP to render a response. For example, in figure 8.8 a Tapestry page contains two links that invoke the Tapestry
application, but the responses are ultimately rendered by a JSP.
Figure 8.8
The Tapestry page used to redirect to one of the JSPs.
370
CHAPTER 8
Advanced techniques
The template for the page is shown in listing 8.11. This page makes use of the
Border component to render the top and bottom of the page. It includes two
links, using the DirectLink component. Clicking those links invokes listener
methods, but the ultimate responses are provided by a pair of JSPs.
Listing 8.11 JSP.html: HTML template for the JSP
<span jwcid="@Border" title="Linking to JSP">
You may:
<ul>
<li><a href="#" jwcid="@DirectLink"
listener="ognl:listeners.displayPageTagDemo">Page Tag Demo</a>:
View a page demonstrating the <tapestry:page> tag</li>
<li><a href="#" jwcid="@DirectLink"
listener="ognl:listeners.displayExternalTagDemo">External
Tag Demo</a>:
View a page demonstrating the <tapestry:external-url>
tag</li>
</ul>
</span>
As you’ve seen before, normally when a DirectLink component is triggered by a
request, the default behavior (after invoking the component’s listener method) is
to render the page containing the component. Alternately, a different Tapestry
page may be activated to render the response. As before, with the banner service,
we’ll use the RedirectException to sidestep this ordinary behavior and instead
redirect to another URL:
public void displayPageTagDemo(IRequestCycle cycle)
{
throw new RedirectException("PageTagDemo.jsp");
}
public void displayExternalTagDemo(IRequestCycle cycle)
{
throw new RedirectException("ExternalTagDemo.jsp");
}
In this case, we are specifying a local resource rather than a complete URL. Tapestry will not send a redirect to the client web browser (as it did with the banner ads
example). Instead, it will instruct the Servlet API to forward the current request to
371
the named resource.3 In this case, the resource is a JSP, but it could be the path of
any servlet, or even a static resource, such as an image or an HTML file.
Relative paths are relative to the servlet. In this case, the two JSPs are in the
context root directory, along with the HTML templates. Absolute paths, starting
with a leading slash, are relative to the context.
These two examples represent some unnecessary complexity; the HTML template could just as easily include <a href="PageTagDemo.jsp"> and <a href=
"ExternalTagDemo.jsp">. However, the same technique works in more complicated cases, such as handling a form submission, or when it is necessary to perform some kind of business operation before redirecting to the JSP.
That’s the easy side of the equation; the tough side is getting a JSP to call into
a Tapestry application.
8.3.2 Linking JSPs to Tapestry pages
Linking from a JSP back into the Tapestry application involves creating a link with
the correct URL. The application URLs used by Tapestry have a very regular format, as we discussed in chapter 7. For example, a JSP could include a link such as
<a href="/app?service=page/Home"> to create a link to the Tapestry Home page.
Despite this regularity, hard-coding links in this way is not advisable:
3
■
A static URL like this is not properly encoded;4 a client that does not support HTTP cookies will lose its HttpSession.
■
If the target page is variable, a Java scriptlet will be needed to construct
the correct URL.
■
If a new release of the framework changes the encoding system used for
application URL s, then the JSP will generate invalid URL s (just such a
change occurred between releases 2.3 and 3.0).
■
If the application has overridden the default engine service implementations (either to encode additional information or to control the format of
the URL in an application-specific way), the JSP may not encode the information correctly.
Specifically, it obtains the RequestDispatcher object for the resource, and then invokes the forward() method on it.
4
All application URLs provided to the client web browser should be encoded using the HttpServletResponse method encodeURL(). This method ensures that the HttpSession ID is encoded into the
URL when the client doesn’t allow cookies. Tapestry services automatically invoke this method as
needed.
372
CHAPTER 8
Advanced techniques
Figure 8.9 The HTML page displayed here was rendered by a JSP and uses the Tapestry tag library
to render the two links back into the Tapestry application.
In short, there’s no substitute for having a Tapestry engine service construct the
URL for that service. That’s what the framework does internally, and it is what we
want here, even though a JSP is responsible for rendering the response. It isn’t
possible to drop a Tapestry component into a JSP, but it is possible to make use of
a close cousin to a Tapestry component: a custom JSP tag.
The Tapestry framework includes a custom JSP tag library expressly for this
purpose. The library allows you to easily create links to pages, leveraging the
Tapestry page and external services. Figure 8.9 illustrates a JSP that makes use of
two of the tags from the library to create links to pages.
The first custom tag is used to create an <a> hyperlink element that links to a
particular page. The second tag creates just the URL portion and is incorporated
into a client-side script that raises the window into a pop-up window. The source
for this page is shown in listing 8.12.
Listing 8.12 PageTagDemo.jsp: JSP using the page and page-url tags
<%@ taglib
uri="http://jakarta.apache.org/tapestry/tld/tapestry_1_0.tld"
prefix="tapestry" %>
b
<html>
<head>
<title>Tapestry JSP Page</title>
</head>
<body>
<script>
<!-function raiseTarget(targetURL)
{
var width = 200;
var height = 200;
var screenX = Math.floor((screen.width)/2)
- Math.floor(width/2);
var screenY = Math.floor((screen.height)/2)
- Math.floor(height/2) - 20;
var features =
"toolbar=no," +
"scrollbars=no," +
"status=no," +
"top=" + screenY + "," +
"left=" + screenX + "," +
"screenX=" + screenX + "," +
"screenY=" + screenY + "," +
"width=" + width + "," +
"height=" + height;
window.open(targetURL, "TargetPage", features).focus();
}
-->
</script>
Two examples of invoking Tapestry pages from a JSP
using the tag library:
<ul>
<li>Using <tapestry:page> to
<tapestry:page page="Target">display the
target page</tapestry:page>
</li>
<li>Using <tapestry:page-url> to
<a href="javascript:raiseTarget(
'<tapestry:page-url page="Target"/>');">raise the
target page in a new window</a>
</li>
</ul>
c
</body>
</html>
d
373
374
CHAPTER 8
Advanced techniques
b
c
d
This code declares the use of the Tapestry tag library within this JSP, using the
prefix tapestry: for the tags provided by the library.
The tapestry:page tag is very similar to the PageLink component; it creates a
link that will reference a Tapestry page.
The tapestry:page-url tag writes out just the URL for a Tapestry page; it is
meant for incorporation into JavaScript, as shown here.
Before we use these custom JSP tags, let’s see how to declare a tag library for use
within a JSP.
Declaring the tag library
Every JSP page that makes use of a tag library must declare the tag library using
a JSP taglib directive, which is shown at the start of the listing. This declaration
identifies the taglib in terms of a resource URI (Uniform Resource Identifier)
and indicates the prefix to be used within the JSP to reference tags defined by the
library. The JSP engine doesn’t look for the named resource on the Internet
(which is fortunate since it does not exist); the resource URI is just used as a key
to indirectly locate the tag library descriptor.
A tag library descriptor is an XML file that defines the tags in a library. For each
tag, it further defines the parameters allowed for the tag. The descriptor for
each tag library used in the application must be included somewhere inside the
web application archive.
The web.xml deployment descriptor matches taglib resource URIs to tag
library descriptors. For the example application, the web.xml deployment
descriptor includes the following elements:
<taglib>
<taglib-uri>
http://jakarta.apache.org/tapestry/tld/tapestry_1_0.tld
</taglib-uri>
<taglib-location>
/WEB-INF/lib/tapestry-3.0.jar
</taglib-location>
</taglib>
This declares the tag library to be packaged as part of the Tapestry framework
JAR . The tag library descriptor is stored inside the Tapestry framework JAR as
the file META-INF/taglib.tld. For this to work, the Tapestry framework must be
stored in the web application archive rather than on the general classpath.
375
Using the page tag
The most basic of the Tapestry custom JSP tags is the page tag, which is very
much like a PageLink component:
<tapestry:page page="Target"> . . . </tapestry:page>
The tapestry: prefix matches the taglib directive previously discussed (it is
often the case that a single JSP makes use of many taglibs, each with its own prefix). Where Tapestry components have parameters, JSP tags have attributes.
Table 8.2 lists the attributes for the page tag.
Table 8.2
Attributes for the Tapestry page tag
Attribute
Required
Description
page
Yes
The Tapestry page we want to link to
servlet
No
The path to the Tapestry servlet; defaults to /app
styleClass
No
If specified, becomes the HTML class attribute
JSP tags don’t have the equivalent of Tapestry’s informal component parameters.
Because it is common to want to specify the CSS class for the <a> tag rendered by
the page tag, an optional attribute for that purpose, styleClass, is provided.
Sometimes, you don’t want the entire <a> tag and just need the application
URL generated by the service. On the example page, the second link doesn’t
directly trigger a request; it invokes a JavaScript function to raise the Target page
in a new window. An <a> element is used that incorporates a page-url tag:
<a href="javascript:raiseTarget(
'<tapestry:page-url page="Target"/>');"> . . .</a>
Unlike the page tag, which wraps around a portion of the JSP, the page-url tag is
empty (much like a Tapestry component that discards its body). The tag is
replaced by the application URL for the named page. The page-url tag accepts
the same page and servlet attributes as the page tag.
8.3.3 Submitting JSP forms into Tapestry
There are times when a JSP will contain a form that must be processed by Tapestry. It isn’t possible to leverage the normal Tapestry form submission mechanism—that’s based on a Form component and form element components within
a Tapestry page, things that can’t exist within a JSP. Instead, the JSP form submission must be processed in the traditional servlets way: by obtaining values for
query parameters.
376
CHAPTER 8
Advanced techniques
Figure 8.10
A JSP containing a simple form. The form will submit to an external page.
But what can a JSP submit to? As before, the level of granularity is an entire Tapestry page, not a component within the page. How could the JSP know the ID of
a component, especially if the component is anonymous? This is where the external service comes in: It provides a hook, in the form of the activateExternalPage() method, where we can process the form submission.
Figure 8.10 shows such a page. It contains a simple form into which you can
enter a name. The form submits into a Tapestry page that displays the name
entered in the JSP-based form.
NOTE
Don’t be confused by the URL displayed in the address field of figure 8.10.
Referring back to chapter 7, you can see that the URL in the window corresponds to clicking some DirectLink within the Tapestry page named
JSP (the page shown in figure 8.9). As you’ve seen, the listener for that
component performed a server-side redirect to the JSP (which is named
ExternalTagDemo.jsp). Because this was a server-side (not client-side)
redirect, the address field still displays the URL used to trigger that DirectLink component.
This JSP builds the <form> around a Tapestry application URL provided by an
external-url tag. The source for the JSP is shown in listing 8.13.
377
Listing 8.13 ExternalTagDemo.jsp: JSP using the external-url tag
<%@ taglib
uri="http://jakarta.apache.org/tapestry/tld/tapestry_1_0.tld"
prefix="tapestry" %>
<html>
<head>
<title>Tapestry JSP Page</title>
</head>
<body>
<form
action="<tapestry:external-url page="JSPForm"/>"
method="POST">
Enter your name:
<input type="text" name="userName" maxlength="100" size="40"/>
<br/>
</form>
</body>
</html>
The example in listing 8.13 resembles the previous one. It starts with a taglib
declaration for the Tapestry tag library. The two key points are as follows:
■
The action URL for the form is provided by the external-url tag.
■
The method for the form is POST, not GET.
Why POST? The URL provided by the external-url tag includes a query parameter, service . When using GET, the client web browser can omit this query
parameter in favor of the query parameters provided in the form submission.
Using the POST method, the query parameters in the action and from within the
form are properly integrated.
On the flip side, when the form is submitted, the JSPForm Tapestry page is
activated. The source for this class is shown in listing 8.14.
Listing 8.14 JSPForm.java: Java class for the JSPForm page
package examples;
import org.apache.tapestry.IExternalPage;
378
CHAPTER 8
Advanced techniques
public abstract class JSPForm extends BasePage
implements IExternalPage
{
protected abstract void setUserName(String userName);
public void activateExternalPage(Object[] parameters,
{
String userName =
cycle.getRequestContext().getParameter("userName");
setUserName(userName);
}
}
The activateExternalPage() method ends up with all the responsibility for processing the form, much like a traditional servlet. Here, the userName query
parameter is extracted and stored into the matching userName page property,
where it will be available to the page’s HTML template when the page renders.
Using service parameters with the external-url tag
As with the ExternalLink component, it is possible to encode service parameters
when using the external-url tag. This is one way to pass additional information
along with the form submission (the other is to make use of hidden form fields).
Table 8.3 lists the attributes that may be used with the external-url tag. The
last attribute, parameters, is used to specify the service parameters. This value
can be either a single string or (by prefixing the string with ognl:) an OGNL
expression. The OGNL expression is evaluated in the JSP context; it will have
access to the HttpServletRequest and HttpSession, as well as any beans declared
in the JSP.
Table 8.3
Attributes used with the external-url tag
Attribute
Required
Description
page
Yes
The Tapestry page to link to, which must implement the
IExternalPage interface
servlet
No
The path to the Tapestry servlet; defaults to /app
parameters
No
Used to specify service parameters
Summary
379
A final note: In addition to the external-url tag, there is an external tag that
generates an <a> element, much like the page tag. Like the page tag, the external
tag includes a styleClass attribute.
8.4 Summary
In previous chapters, we discussed how to extend the Tapestry framework from
within, by using the services and components available with the framework, even
to the point of creating new components. That’s just scratching the surface of
Tapestry. Because of its open-ended, object-oriented structure, the framework
has untold avenues for extension. You can bend Tapestry to the needs of your
project—even when your project needs new kinds of services, or complex clientside scripting, or even integration into an existing JavaServer Pages application.
So far, the examples and discussions in this book have concerned isolated
examples, but that’s not how Tapestry is used in real life: It is used to create
working applications. Chapters 9 and 10 describe a real-world Tapestry application and demonstrate how the approaches explained in this and the preceding
chapters are utilized in a real project.
Part 3
Building complete Tapestry
applications
C
hapters 9 and 10 focus on the Virtual Library, a Tapestry-driven example J2EE application. Within the context of a complete application, you’ll be
exposed to many of the design and development issues common to web applications, and you’ll see good examples of how to address those issues.
This chapter covers
■
The Virtual Library application
■
Thinking about application flow
■
Limiting access to parts of the application
383
384
CHAPTER 9
So far in this book we’ve focused on individual components or simple isolated
examples—but that’s not how Tapestry is supposed to be used. New issues arise
when you’re creating a real working application:
■
Who is the user?—Most applications have some concept of users and
logging in, and some aspects of the application change once the user has
been authenticated.
■
Can the user view this page?—Parts of an application may be accessible
only to users who have logged in. Some applications may impose more
restrictions, preventing some users from accessing certain pages.
■
How do I know what to display on this page?—More important, if the page
is redisplayed (for example, if the user logs in and is returned to the same
page afterward), is there server-side state to determine what to display?
In this chapter, we’ll describe the functionality of a demonstration application
included in the Tapestry distribution: the Tapestry Virtual Library. In the next
chapter, we will implement the application.
9.1 Introducing the Virtual Library
Ever notice all those books lining the shelves of your cubicle, as well as all the
cubicles of your neighbors? Ever notice the degree of duplication of books? You
might need a book just for a week or two, and so you buy your own copy of it. It
might be nice to get a copy from the library, but public libraries are probably not
up to date enough with cutting-edge technical books. What if there was a way to
find a copy of a book you needed that just happened to be gathering dust on a
coworker’s shelf? What if there was a way to track where borrowed books are
(which makes the books’ owners more amenable to lending them out)? That’s the
Virtual Library, a kind of match-making application for books.
The Virtual Library is an application, built using Tapestry, that implements
this idea. It tracks books and the people who own and borrow those books. The
Virtual Library includes search capabilities that allow you to find books of interest and records that you’ve borrowed them. It also includes all the necessary
pages for creating and editing a record of your books.
The Tapestry distribution includes the precompiled Enterprise Application
Archive (EAR) containing the presentation layer for the Virtual Library, as well as
the Enterprise JavaBeans (EJBs) that perform database access.
Performing searches
385
This chapter takes you on a tour of all the pages in the application, spotlighting interesting challenges related to the functionality on each page. Chapter 10
describes the implementation of the Virtual Library application.
The Tapestry distribution includes all the necessary directions for configuring
and starting an instance of the open-source JBoss application server to execute
the Virtual Library application. Nearly all the configuration is done automatically using an Ant build script. Once JBoss is up and running, the Virtual Library
application may be accessed using the URL http://localhost/.
9.2 Performing searches
After launching the application, you will be presented with the application’s
search page, shown in figure 9.1. Since you must locate books before you can
borrow them, this page is the central focus of the application. The Search page
includes a form that allows you to search for books by entering a partial title or
author name, the name of the book’s owner (using a drop-down list), or the
name of a publisher (again, using a drop-down list).
All pages in the Virtual Library include a navigational border that identifies
the application and page title along the top, and that provides a navigation
menu linking to different parts of the application along the left edge. Initially,
Figure 9.1 The Search page allows books to be located by title, author, owner, or publisher. The
navigation menu along the left side expands once you execute the search, or after you log into the
application.
386
CHAPTER 9
Figure 9.2 Search results listing matching books. Columns may be sorted using the links and icons
across the top, and the results may be paged through using the controls along the left edge.
the only selections in the navigation menu are Search, My Library, and Login.
Once you log in, more options appear in the navigation menu.
After entering search terms and clicking the Search button, you are presented
with a listing of books matching the search criteria (figure 9.2). The Matching
Books page presents these results in a table format, complete with navigation
options for paging through the results.
The main listing is the title, author, publisher, owner, and current borrower (if
any) for each book. Names of books and people are clickable links leading to
detail pages about the book or user. Recently added books are marked with a
“new” icon. Until you log in, books added in the previous month are marked
new; but after you log in, the application knows when your previous visit
occurred, and only books added since then are marked new.
The far-right column, Borrow, contains buttons for borrowing books. The
column starts empty, since only authenticated users may borrow books. After you
log in, icons will appear for books you may borrow—that is, books you aren’t
already in possession of. In addition, a book owner may mark a book as nonlendable; such a book will appear in the listing, but no borrow icon will ever appear
for it.
Performing searches
387
Figure 9.3
Clicking the column name or
the sort icon toggles between
ascending and descending
sort order. Clicking a different
column title re-sorts the table
on that column instead.
9.2.1 Changing the table sort order
The columns may also be sorted; initially the listing is sorted by book title (in
ascending order), which is indicated by the triangle icon in the Title column. Figure 9.3 shows how you may sort columns using either links or visible icons. Clicking on a column title sorts (or re-sorts) the results according to the content
within that column.
Many pages in the application display lists of books, much like the Matching
Books page. All of these pages allow you to control sort order in the same way.
9.2.2 Paging through the results
Because the book listing may be very large, the Virtual Library pages the results,
showing 15 books per page. When a book listing is displayed, a set of page navigation controls appears along the left side, under the navigation menu. Figure 9.4
shows the detail for this portion of the page.
These controls show the total number of books in the listing and the range
currently showing. Four buttons control navigation: First, Previous, Next, and
Last. The First and Previous buttons are disabled on the first page of the listing,
and the Next and Last buttons are disabled on the last page. Disabled buttons
are drawn “grayed out” to provide additional feedback to you that it is not
appropriate to click the button.
As with sorting, all pages that display lists of books use the same page navigation controls.
Figure 9.4
This control allows navigation through the query
results a page at a time (each page showing 15
books), or lets you jump to the first or last page. The
buttons on the left are disabled since the first page is
showing. The button under the cursor is highlighted.
388
CHAPTER 9
Figure 9.5 The user’s email address is used as a login ID. You may also register on the fly. Return
users will see their email address filled in automatically and have to enter only their password.
9.3 Logging in and registering
You may log into the Virtual Library from any page by clicking the Login link in
the navigation menu. After you log in, you are returned to the page where you
started. The My Library link will also allow you to log in but will send you to the
My Library page instead.
The Login page, shown in figure 9.5, makes use of the validation subsystem
described in chapter 5. Entering an unknown email address or an invalid password redisplays the form with an error message and the invalid fields well
marked, as shown in figure 9.6. For security reasons, even if you entered a password, the Password field is redisplayed empty.
Figure 9.6 The Virtual Library uses the validation subsystem from chapter 5. Errors are
clearly displayed and invalid fields marked.
Borrowing books
389
Figure 9.7 New users can register at any time by providing their name, email, and password. The
password is entered twice to ensure accuracy.
The sample database that comes with the Virtual Library includes three users:
[email protected], [email protected], and [email protected]. All three
use the same password, secret.
New users can register on the fly by clicking the “Register now” link on the
Login page. This displays the Register page, shown in figure 9.7.
After logging in or registering, you are returned to the page where you
started. Because you are now authenticated, borrow icons will appear for books
you can borrow, and additional navigation links are shown, as illustrated in figure 9.8.
9.4 Borrowing books
To borrow a book, click the borrow book icon (it appears in the Matching Book
page, and in the View Person and View Book pages we haven’t discussed yet).
This will update the Virtual Library database, identifying you as the holder of
the book. This is your opportunity to stroll over to the owner’s cubicle and get
the actual, physical book. When borrowing a book, you will be returned to the
Search page, and an informational message will be displayed, as shown in figure 9.9.
390
CHAPTER 9
Figure 9.8 After logging in (or registering), you are returned to the original page. You may now
borrow books, and you have additional navigation options (listed under My Library).
9.5 Getting details about books and persons
Throughout the application, whenever the name of a book or the name of a person appears, it is displayed as a link. Clicking the link directs you to a details
page, showing more information about that book or about that person.
9.5.1 Viewing book details
Figure 9.10 shows an example of the View Book page. The page displays all the
information provided by the book’s owner, including a long description of the
book. The page includes links to the book’s owner and the person currently
holding the book (which may be the book’s owner or another user borrowing the
Getting details about books and persons
Figure 9.9
appears.
After borrowing a book, you are returned to the Search page, and a reminder
Figure 9.10 Clicking the name of a book displays the details about a book, with an icon for
borrowing the book, as well as links to the book’s owner and holder.
391
392
CHAPTER 9
book). In addition, the View Book page includes a borrow icon (but only if the
current user has logged in).
The ability to log into the application on the fly is very handy here. Figure 9.10
shows an example of the View Book page after a user has logged in; it includes a
borrow icon. If you were, instead, to launch the Virtual Library and do a search
without logging in first, there would be no borrow icon. A poorly designed system would require you to log in and then perform the search again before you
could borrow the book. The Virtual Library allows you to log in on the fly and
automatically returns you to the page you were viewing before logging in, at
which point a single additional click will borrow the book.
9.5.2 Viewing a person
Clicking on a person’s name displays the View Person page for that person,
which includes a listing of all the books owned by that person (if any). An example of the View Person page is shown in figure 9.11. As elsewhere, the list of
books can be paged through or re-sorted, and borrow icons appear for all appropriate books, provided the current user has logged in.
Figure 9.11 The View Person page includes a listing of the books owned by the person, with
yet more icons for borrowing books. The book Bitter Java may not be borrowed, because
Bender Robot is the current user and is currently holding the book.
Managing your books
393
The borrow icon does not appear unless the user is logged in. Even for a
logged-in user, the icon does not appear if the current user is already holding
the book, or if the book is not lendable (the owner of a book selects whether
books can be borrowed).
9.6 Managing your books
The My Library page is the starting point for managing the books that you own.
From the My Library page (and by using the menu items listed below My
Library), it is possible to add records for new books, modify existing books’
records, and even give books away to other users of the application.
The My Library page (shown in figure 9.12) shows the entire list of books the
current user owns, with controls to allow you to edit or delete existing books. It is
Figure 9.12
This interface is used to edit or delete books owned by the current user.
394
CHAPTER 9
accessed using the My Library link in the navigation menu. As with other book
listings, the My Library page includes navigation controls for paging through
the list and lets you change the sort order of the list. Each book has a pair of buttons for editing and deleting the book. Unlike the other pages, the book titles in
My Library are not links; this is to remove any ambiguity about what clicking the
link would do (would it show the book detail or edit the book?).
9.6.1 Editing a book
Clicking the edit icon for a book in the My Library page allows the details of a
book to be edited. The Edit Book page, shown in figure 9.13, is a form allowing you to update the details of the book, including the current borrower of
the book.
Figure 9.13 Books may be edited, updating all information, including who is currently borrowing the
book. Publishers may be selected from a list or entered into the text field (but not both).
Managing your books
395
An interesting aspect of this page is the Publisher field. You can either select an
existing publisher from the drop-down list or leave the drop-down blank and
enter the name of a new publisher into the adjacent text field. This input behavior is backed up by client-side JavaScript that enables and disables the text field
based on the selection in the drop-down list. The text field is disabled if the
drop-down list has a nonblank value selected. Selecting the blank entry in the
drop-down list enables the text field and moves the cursor into it.
9.6.2 Deleting a book
Books may be deleted by clicking the delete icon on the My Library page. This
results in a confirmation page, shown in figure 9.14.
9.6.3 Returning books
The Borrowed Books menu item shows a listing of all books the logged-in user is
currently borrowing, as shown in figure 9.15. From here, you can also return
books by clicking the book return icon (in the rightmost column). Like all other
pages that show lists of books, the Borrowed Books page includes page navigation controls and the ability to change the sort order of the list.
Figure 9.14 Clicking a delete icon on the My Library page displays a confirmation page before the
book is actually deleted.
396
CHAPTER 9
Figure 9.15 The Borrowed Books listing shows all books borrowed by the current user. The return
icon in the rightmost column allows books to be returned.
9.6.4 Adding a new book
Before there can be books to edit or delete, books must be added. The Add New
Book page, accessed via the Add New Book item in the navigation menu, is the
means for adding new books. This page is shown in figure 9.16.
As with the Edit Book page, the Publisher drop-down list is exclusive with the
adjacent text field.
9.6.5 Editing your profile
The Edit Profile page, shown in figure 9.17, is used to edit your name, email
address, and password. It is accessed using the Edit Profile item in the navigation menu.
9.6.6 Giving away books
Sometimes you’ll want to give away books. You may no longer need the book, or
you may just be clearing out some shelf space. One way to move ownership of a
Managing your books
Figure 9.16 Adding a new book is much like editing an existing book, except that the
borrower cannot be edited.
Figure 9.17
password.
The Edit Profile page allows users to update their name, email address, and
397
398
CHAPTER 9
book is to delete it from your My Library and let the new owner add it to his or
her My Library. A better way is simply to shift the ownership of the book without
otherwise changing it. This is accomplished using the Give Away Books item on
the navigation menu.
The Give Away Books page, shown in figure 9.18, lets you select one or more
books and designate a person to give those books to. The ownership of the books
is then immediately changed over to the new owner.
Obviously, we’re using a Palette component (discussed in chapter 6) to select
the books to transfer, but notice how we’ve customized its configuration. Normally, the column titles are Available (on the left) and Selected (on the right). On
this page, the left column is labeled “Books you currently own,” and the right
column includes a drop-down list for selecting the target person (to transfer
books to).
That covers all the pages that may be reached by ordinary users. The next
section covers the remaining pages accessible to Virtual Library administrators.
Figure 9.18 The Give Away Books page allows the ownership of books to be changed. A target user
is selected, and the Palette component is used to identify which books are to be transferred.
399
9.7 Administering the Virtual Library
As a demonstration application, the Virtual Library extends past functionality
that ordinary users will see. It includes additional functionality needed to administer the application. A less complete application might omit this functionality,
forcing administrators to perform these operations directly against the database.
Instead, the Virtual Library seamlessly integrates this functionality into the overall application.
The administrative pages are protected; if you attempt to access them before
logging in (or after your session expires), you will be forced to log in. If you don’t
have administrative privileges, you’ll be returned to the Search page with an
error message.
If you do have administrative privileges, you will see additional menu items
in the navigational menu. Administrators can
■
■
■
■
■
Delete existing users
Override passwords for users
Add or remove administrative privilege for users
Manage the list of publishers
Transfer ownership of books between users
Of these, the most common activities relate to managing users, so we’ll start there.
9.7.1 Editing users
The chief job of an administrator is to manage the other users of the Virtual
Library. This is done through the Edit Users item in the navigation menu, which
displays a list of all users in the database (as shown in figure 9.19).
To test the administrative pages, you may log in as user ringbearer@
bagend.shire, using the password secret. Three new options will appear in the
navigation menu: Edit Users, Edit Publishers, and Transfer Books.
The page allows data about other users to be changed; the current user is
omitted from the list. This ensures that administrators do not accidentally delete
themselves or remove their own administrator privilege. User names and email
addresses cannot be changed, but users can have administrator access granted or
revoked, can be locked out of the system entirely, can have their password
changed, and can be deleted outright.
When a user is deleted, any books owned by the user are reassigned to the
current user, the administrator who performed the deletion. Additionally, any
books borrowed by a deleted user are returned to their owners.
400
CHAPTER 9
Figure 9.19 The Edit Users page allows an administrator to modify or delete the users in the
system. The current user ([email protected] in this figure) does not appear in the list.
9.7.2 Editing publishers
When users add or edit books, they often select a known publisher for the book
from the drop-down list. Users may also create new publishers as needed. Of
course, they won’t always get it right, and the Edit Publishers page, shown in figure 9.20, allows an administrator to clean things up. The page is accessed using
the Edit Publishers item on the navigation menu.
The Edit Publishers page allows all the known publishers to be edited in a
single place. In the Virtual Library, a publisher is simply a name (the point is to
populate the drop-down list with a reasonable value). The Edit Publishers page
enables all the publishers be renamed or deleted in a single operation.
9.7.3 Transferring books
The final operation for administrators is to forcibly transfer book ownership
between users. This is much like the Give Away Books page, except the administrator can choose any two users to transfer between. This operation takes the
form of a two-page wizard, accessed using the Transfer Books item in the navigation menu. The first page of the wizard is shown in figure 9.21.
Figure 9.20
outright.
The complete list of publishers is displayed. Publishers may be renamed or deleted
Figure 9.21 The Transfer Books wizard starts by having the adminstrator select the two users to
move books between.
401
402
CHAPTER 9
Figure 9.22 After the administrator selects two users, the second step is to choose the books to
be transferred to the target user.
The first step is to select the two users to move books between. The source user
must own some books. The target user can be anyone but the source user. Once the
two users are selected, the second page in the wizard (see figure 9.22) is displayed.
9.8 Summary
The Virtual Library is a small, simple application with only three entities (books,
persons, and publishers); even so, you’ve seen that a proper approach to this
application is moderately complex. There’s a lot of polish to the application,
with features such as on-the-fly login and registration (that remembers where
you started) and consistent column sorting and page navigation, and even little
features such as rollover buttons. In addition, we’ve included a suite of administrative pages that will keep the owner of the Virtual Library from having to muck
about with the underlying SQL database.
Chapter 10 continues with the Virtual Library, providing details on how each
page and component is implemented.
Implementing a Tapestry
application
This chapter covers
■
Dividing the application into layers
■
Combining Tapestry with Enterprise JavaBeans
■
■
Adding security checks to pages
■
Deploying an enterprise application
403
404
CHAPTER 10
In the previous chapter, we described the Virtual Library, a demonstration
application included with the Tapestry distribution. We covered quite a bit of
functionality, some of which is specific to individual pages, while other functionality is distributed throughout the entire application. In this chapter, we’ll get
down to the nuts and bolts of implementing the Virtual Library. Along the way,
we’ll investigate how to best utilize the facilities of Java 2 Enterprise Edition
(J2EE) to support the application, and how to best organize our code for robustness and maintainability.
The source code for the Virtual Library is part of the main Tapestry distribution; the code for the Enterprise JavaBeans (EJBs) is in the examples/VlibBeans
directory; and the code for the Tapestry portion of the application is in examples/Vlib. Appendix B tells you how to obtain the Tapestry distribution.
10.1 Looking at the application layers
Because it’s a J2EE application, the Virtual Library is split into two distinct sections: a presentation layer and an application layer, as shown in figure 10.1. The
presentation layer is the user-facing aspect of the application, and is the portion
constructed using Tapestry. The application layer consists of the session and
entity EJBs that, ultimately, interact with the database.
The application layer for the Virtual Library consists of six EJBs (listed in
table 10.1): three entity beans using container-managed persistence, and three
session beans. The presentation layer will only access two of those beans: BookQuery and Operations.
Figure 10.1 The presentation layer, the Tapestry portion of the application, is separate from the
application layer.
Looking at the application layers
Table 10.1
405
EJBs used to manage interactions with the database
Bean
Usage
Book
Entity bean for books.
BookQuery
Stateful session bean for executing queries about books. Creates and executes
dynamic JDBC queries and stores results.
KeyAllocator
Stateless session bean; allocates primary keys for entity beans.
Operations
Stateless session bean for performing most operations, including creating, reading,
and updating entity beans (using value objects).
Person
Entity bean for persons (a.k.a. users).
Publisher
Entity bean for publishers.
All database access, without exception, occurs in the application layer, as either
direct Java Database Connectivity (JDBC) calls in the Operations, BookQuery, or
KeyAllocator bean, or as container-managed persistence in the three entity
beans. In a production environment, the presentation and application layers will
be operating within the same Java Virtual Machine (JVM). Most application servers (including JBoss and WebLogic) will recognize this configuration and have
the different objects communicate directly, without utilizing Remote Method
Invocation (RMI). RMI is intended for communications between different processes often on different servers, so avoiding that overhead provides a great performance boost.
In a development environment, things are different. During development,
the application layer changes far less often than the presentation layer. When
was the last time you tweaked an EJB because it “didn’t look right”? Here’s where
the enforced discipline of developing a J2EE application, with real separation
between the layers, actually pays off: The application layer can run in one process, just as it would in a production environment, while the presentation layer
runs in another process, executing within your IDE. Important advantages to
structuring your development environment this way include the following:
■
Debugging the presentation layer is much faster and easier than debugging the application server remotely, especially when using an IDE that
supports “hot code replace” (changing methods while the application is
still running), such as Eclipse.
■
The presentation layer may be stopped and restarted in a manner of a few
seconds. Even a lightweight application server such as JBoss can take at
least a minute, often much longer, to shut down and restart.
406
CHAPTER 10
■
You can work on files directly within your development workspace, without
constantly having to repackage and redeploy your web application.
■
You will discover early any problems with object serialization (of any value
objects that move between the two layers). Operating within a single JVM
can mask these problems (since RMI will not be used).
10.2 Organizing EJB access
The Virtual Library makes use of Tapestry features to organize the application
sensibly. This organization takes the form of specific responsibilities for the
engine, the Visit object, and the Global object. The Global object is similar to
the Visit object but is shared by all sessions. Figure 10.2 shows how these central
objects are stored on the server. The engine and the Visit object are stored
within the HttpSession as attributes; there will be many engine and Visit
instances, one pair for each concurrent user of the application.
The Global object, on the other hand, is global to (and shared by) all engine
instances, stored as a ServletContext attribute. In the Virtual Library, the Global
object is responsible for performing Java Naming and Directory Interface (JNDI)
lookups on behalf of the engine, locating home interfaces for the Operations
and BookQuery EJBs. This is an excellent use for the Global object, since it can
cache references to the EJB home interfaces. JNDI lookups of home interfaces
are a notoriously slow operation, so caching the results for quick access is an
effective performance enhancement.
As you’ll see shortly, locating these lookup operations centrally leads to not
only a better-performing application but also a more robust one. Having a single,
authoritative location for JNDI lookups means that there will be a single, authoritative approach to handling JNDI lookup errors. It also provides a single place
to manage recovery from RemoteExceptions. A traditional servlet application
can have JNDI lookups and cached home interface references scattered across
Figure 10.2 The engine is responsible for common EJB operations. The Visit object is responsible
for authentication. The Global object, which is responsible for JNDI lookup, is a singleton, shared
by all engines.
407
many servlets and JSPs, but Tapestry makes it easy to centrally locate these
important operations.
10.2.1 Handling authentication
The Visit object is responsible for authentication. It simply tracks the ID of the
currently logged-in user. This ID is an Integer, the primary key type of all three
entities used by the Virtual Library (Person, Book, and Publisher). Additionally,
the Visit object retains the matching instance of Person, so that full details about
the user (such as name, email address, and administrative privilege) can be made
available to the application.
10.2.2 Accessing Enterprise JavaBeans
The majority of interaction between the two layers involves the Operations bean.
This is a stateless session bean, and it consists of a number of methods for creating, reading, and updating back-end data. The application uses a subclass of the
BaseEngine class. The subclass has extensions for obtaining a reference to the
Operations bean’s remote interface as needed, as well as implementing several of
the most common operations, such as reading a Person value object corresponding to a person ID.
As with JNDI lookups in the Global object, having operations implemented in
the engine centralizes handling of, and recovery from, remote exceptions. This
leads to much greater application stability, especially during redeployment of the
application. In the code excerpts that follow, you’ll see how the engine provides
support for retry loops used in every location where remote object access occurs.
10.2.3 Tracking user identity with the Visit object
The specific requirements for the Visit object in the Virtual Library are that it
track the following:
■
Who the logged-in user is (if any)
■
The last time the user accessed the system (information used to identify
which books are “new”)
Listing 10.1 shows how this class is implemented.
Listing 10.1 Visit.java: Java class for the Virtual Library Visit object
package org.apache.tapestry.vlib;
import java.sql.Timestamp;
408
CHAPTER 10
import org.apache.tapestry.vlib.ejb.Person;
public class Visit implements Serializable
{
private static final long serialVersionUID =
8589862098677603655L;
private transient Person _user;
private Integer _userId;
private Timestamp _lastAccess;
b
public Timestamp getLastAccess()
{
return _lastAccess;
}
public Person getUser(IRequestCycle cycle)
{
if (_user != null)
return _user;
if (_userId == null)
return null;
VirtualLibraryEngine vengine =
(VirtualLibraryEngine)cycle.getEngine();
c
_user = vengine.readPerson(_userId);
return _user;
}
public Integer getUserId()
{
return _userId;
}
public void setUser(Person value)
{
_lastAccess = null;
_user = value;
_userId = null;
if (_user == null)
return;
_userId = _user.getId();
_lastAccess = _user.getLastAccess();
}
409
public boolean isUserLoggedIn()
{
return _userId != null;
}
public boolean isUserLoggedOut()
{
return _userId == null;
}
public boolean isLoggedInUser(Integer id)
{
if (_userId == null)
return false;
return _userId.equals(id);
}
public void clearCache()
{
_user = null;
}
}
b
c
The _user instance variable stores the Person data object for the logged-in user.
Because the field is transient, its value may be lost at any time due to clustering.
Whenever the _user variable is null, it can be reacquired from the engine (using
the VirtualLibraryEngine subclass).
The Visit object is stored in the HttpSession as a session attribute after most
requests. On any request where the Visit object is accessed, the framework
makes the assumption that the Visit has been changed.1 This triggers the Tapestry ApplicationServlet to store the engine (and with it, the Visit) in the
HttpSession at the end of the request, after the response page has been sent back
to the client.
In a nonclustered environment, this has little impact because the HttpSession
is just a glorified java.util.Map; storing attributes in the session doesn’t trigger
any additional behavior. In a clustered environment, the engine and, with it, the
Visit may be serialized and copied to a backup server within the cluster. It’s in
1
Since the engine doesn’t know anything about the Visit, it is forced to assume that any access of the
Visit is a potential change to the Visit that must be propagated throughout the application server
cluster.
410
CHAPTER 10
the best interests of general application performance for the Visit object to serialize efficiently.2
The Visit object declares a static serialVersionUID constant. Declaring this
as a static variable means that the JVM does not have to compute this value at
runtime, leading to a modest efficiency improvement. A better improvement is
to implement the java.io.Externalizable interface (instead of the java.io.
Serializable interface). This allows you to write more efficient methods to save
and restore the state of the object. This is an example of where runtime performance profiling would be used: to see just how much time is being spent serializing the Visit object. Based on that data, a decision about the merits of
expending effort on serialization of the Visit can be made.
The Visit object tracks the logged-in user’s ID, but also the Person object
representing the user. The latter is the value object for the Person entity and is
needed in the presentation layer to display the user’s name or to determine the
user’s administrative privilege. These user ID and Person instances are stored in
two separate instance variables. The user’s ID is stored persistently, but the Person object is stored transiently. This means that the Person object, within the
Visit, does not have to be serialized when the Visit is itself serialized, an obvious optimization—but only if the Person object can be restored when needed.
This on-demand restoration is handled in the getUser() method. Rather
than being a pure accessor method (taking no arguments), this method requires
that the IRequestCycle object be passed in. If the Person object has been lost,
this method can use the cycle to obtain the VirtualLibraryEngine instance for
the application, and use the engine to reread the Person.
You’ll be seeing this pattern in different guises repeatedly: store the minimum
persistent state possible and recover the full state only as needed. The Tapestry
framework provides the necessary structure for maintaining the minimal state
and recovering the full state.
10.2.4 Understanding page inheritance
The inheritance hierarchy for pages is shown in figures 10.3 and 10.4. Figure 10.3
shows interfaces and base classes, and figure 10.4 shows how each page in the Virtual Library inherits from or extends from those interfaces and base classes.
At first glance there is an awfully large number of classes, but this is natural
for a Tapestry application: Each page has its own class, even though in most
2
The engine implements the java.io.Externalizable interface for efficiency reasons.
411
Figure 10.3
The Virtual Library includes
three interfaces (IActive,
IErrorProperty, and
IMessageProperty) and
three base classes
(ActivatePage, AdminPage,
and Protected) from which
the actual pages extend.
cases, the class is only a few lines long. The presentation layer accounts for
approximately 1,850 lines of code, excluding comments. This is a surprisingly
small number of lines, considering the amount of functionality present in the
Virtual Library application. The application layer (the EJB implementations and
value objects), at about 1,250 lines, is slightly smaller. Of course, a significant
portion of the presentation layer is in non-Java artifacts: HTML templates, page
specifications, and component specifications.
The Virtual Library includes base classes and interfaces that define additional
logic. These classes and interfaces are primarily concerned with preventing unauthorized users from seeing pages, at least until they are authenticated. In addition,
the Java classes for the ViewPerson and ViewBook pages implement the IExternalPage interface. This is to allow links to persons and books to be stored as bookmarks. Table 10.2 lists the base classes and interfaces used by the application.
Table 10.2 Base classes and interfaces used to define and implement behavior common to many
application pages
Class/Interface
Usage
AdminPage
Restricts access to the page to users with administrative access.
Protected
Restricts access to the page to logged-in users, forcing an on-the-fly login as
necessary.
412
CHAPTER 10
Table 10.2 Base classes and interfaces used to define and implement behavior common to many
application pages (continued)
Class/Interface
Usage
IActivate
Defines a page that needs an extra initialization phase before it initially renders.
IErrorProperty
Defines a string property named error. Matches a property specification in
the page specification.
IMessageProperty
Defines a string property named message. Matches a property specification in
the page specification.
Now that we’ve established the general framework of the application, we can
start to drill down into the pages and components of the Virtual Library.
Figure 10.4 The 17 pages of the application extend from framework base classes,
such as BasePage, or from application-specific base classes, such as
ActivatePage.
413
10.3 Implementing the Search page
In the Virtual Library, the page used for searching is the Home page (it is
labeled Search in the user interface but is the Home page nonetheless). The function of this page is to collect user input (the title, author, and so forth), execute
the query, and use the BookMatches page to display the results. The Home page
is, in many ways, the simplest page in the application. Refer to figure 9.1 to see
what the Home page looks like.
The majority of the template for the Home page (listing 10.2) is the Form
component and the <table> element used to control its layout. At runtime, the
majority of the HTML is produced inside the Border component.
Listing 10.2 Home.html: HTML template for the Home page
<html jwcid="$content$">
<body jwcid="@Border">
<p>
Use this form to locate books stored in the Virtual Library.
When you find a book you want,
<a href="#" jwcid="@PageLink" page="Login">login</a> and borrow
it  just don't forget to give it back eventually!
</p>
<span jwcid="@ShowError"/>
<span jwcid="@ShowMessage"/>
<form jwcid="@Form" listener="ognl:listeners.search"
stateful="ognl:false">
<tr>
<th>Title</th>
<td><input type="text" jwcid="inputTitle@TextField"
value="ognl:title" size="30" maxlength="100"/>
</td>
</tr>
<tr>
<th>Author</th>
<td><input type="text" jwcid="inputAuthor@TextField"
value="ognl:author" size ="30" maxlength="100"/>
</td>
</tr>
<tr>
<th>Book Owner</th>
<td><select jwcid="inputOwner@PropertySelection"
value="ognl:ownerId"
model="ognl:engine.buildPersonModel(true)"/></td>
414
CHAPTER 10
</tr>
<tr>
<th>Publisher</th>
<td><select jwcid="inputPublisher@PropertySelection"
value="ognl:publisherId"
model="ognl:engine.publisherModel"/>
</td>
</tr>
<tr>
<td></td>
<td>
<input type="image" src="images/search.png" width="46"
height="20" border="0" alt="Search"/>
</td>
</tr>
</table>
</form>
</body>
</html>
The Form component for the search form specifies a value of false for the Form’s
stateful parameter. By default, the stateful parameter is true, but setting this
parameter to false allows the application to recover more gracefully from an
expired HttpSession. If a user logs into the system and returns to the Home
page, and then lets the session expire (by waiting 15 minutes or so) before submitting the search form, the application will still be able to perform the search,
even though a new HttpSession will be created in the process and the user will
no longer be logged in.
Leaving the stateful parameter in its default value, true, changes this scenario. When submitting the form after the session expires, Tapestry will recognize
that the session is stale and throw a StaleSessionException. The VirtualLibraryEngine handles the StaleSessionException by returning the user to the Home
page with a message about being logged out.3 Although better than displaying the
StaleSession page, it is still less than ideal, and is quite easy to avoid. There are
times when it is better to leave the stateful parameter as true—for example, in
forms that require the user to be logged in for the forms to operate.
3
The default behavior in this case is normally to display the StaleSession page; the Virtual Library uses
its own subclass of BaseEngine and overrides the engine method handleStaleSessionException()
to instead redirect the user to the Home page with an error message.
415
10.3.1 Identifying application-specific components
The first thing you’ll notice in the Home page template is that there isn’t any
HTML for the title bar that spans the top of each page, or for the navigation
menu that runs down the left side. These common fixtures are entirely produced
by the Border component, an application-specific component we’ll discuss in a
bit. The Border component is referenced in the <body> tag of the template.
In fact, all the templates in the application are minimal; they don’t make
more than a small effort to maintain WYSIWYG preview. The Home page’s
HTML template is a valid HTML file, but it does not reference a stylesheet. The
end result is only a limited ability to preview the application using an HTML editor. Figure 10.5 shows what such a preview looks like.
Figure 10.5 Previewing the Home page in a WYSIWYG editor is only a limited success.
Navigational borders and stylesheet information are missing. Compare this to figure 9.1, taken
from the live application.
416
CHAPTER 10
Two other application-specific components are used in the template: ShowError
and ShowMessage. They are used to display errors or informational messages.
Figure 9.6 shows an example of the ShowError component (on the Login page),
and figure 9.9 shows an example of the ShowMessage component.
These components don’t take any parameters; they expect that the page
implements an error property (for ShowError) or a message property (for ShowMessage). If you refer to figure 10.3, you’ll see that there are two interfaces,
IErrorProperty and IMessageProperty. These two interfaces define the accessor
methods for the two properties. Java page classes that implement one or both
interfaces include matching property specifications for the properties in the
pages’ specifications.
10.3.2 Referencing the engine
Two components in the template reference the engine: the PropertySelection component for selecting a book owner, and the PropertySelection component for
selecting a publisher. Creating a custom subclass of BaseEngine for your application
is a great way to centrally locate common business logic and utility methods in a
way that is easily accessible from any page or any component. Here, the engine is
being used to create property selection models for persons and publishers.
The inputOwner component uses the engine to build a model of all the users
in the system:
<select jwcid="inputOwner@PropertySelection"
value="ognl:ownerId"
model="ognl:engine.buildPersonModel(true)"/>
The engine’s buildPersonModel() method invokes the Operations bean to obtain
this list of persons. It constructs and returns an IPropertySelectionModel
instance. The argument, specified as true, indicates that an initial empty element
should be included in the model. This is necessary in the user interface, because
the user may not want to filter results by book owner at all.
Likewise, the inputPublisher component uses the engine to build a model for
selecting a publisher:
<select jwcid="inputPublisher@PropertySelection"
value="ognl:publisherId"
model="ognl:engine.publisherModel"/>
Here, the engine provides a simple property that is an IPropertySelectionModel
of all the publishers currently in the database. This model always includes an
empty element (so that the user may choose not to filter by publisher).
417
10.3.3 Specifying the page class and properties
The meat of the page is the Form and form element components that collect the
search terms from the user. Each of these form element components updates a property of the page, a property created at runtime from a <property-specification> in
the Home page’s page specification (see listing 10.3).
Listing 10.3 Home.page: specification for the Home page
<?xml version="1.0" encoding="UTF-8"?>
<page-specification class="org.apache.tapestry.vlib.pages.Home">
<property name="page-type" value="Search"/>
<property-specification name="error"
<property-specification name="message"
Specifies page type for
Border component
Specifies error property for
IErrorProperty interface
Specifies message property for
IMessageProperty interface
<property-specification name="title"
<property-specification name="author"
<property-specification name="ownerId"
type="java.lang.Integer"/>
<property-specification name="publisherId"
type="java.lang.Integer"/>
Specifies
properties for
user input
Part of the page specification is a piece of metadata needed by the Border component. Metadata is extra information that can be attached to a page or component
specification. Each metadata property is in the form of a key-value pair (keys and
values are strings). This information is read from the specification by the framework and is available to the application at runtime but is not used by Tapestry.
Instead, it’s used only by the application. Page and component specifications (as
well as application and library specifications and many other elements within the
specifications) support metadata in the form of <property> elements. A specification may have any number of metadata properties, with each value identified by a
unique key. In this case, the metadata property is named page-type.
418
CHAPTER 10
This metadata property is read by Border component; this is how it knows
which page to display in the title bar and which images to use when displaying
the navigation menu. Not all pages need it—just “rule-breakers” like the Home
page (which needs to be labeled Search). When we discuss the Border component
a bit later, you’ll see exactly how this metadata property is used.
The rest of the page specification consists of property specifications used on
the page. Because this page uses the ShowError component, the class implements the IErrorProperty interface, and a <property-specification> for the
error property is included. Likewise, the ShowMessage component requires the
IMessageProperty interface and a specification for the message property.
10.3.4 Performing searches
All of these templates, components, and specifications come together in the Home
page’s Java class. The class includes abstract accessor methods for the four properties supplied by the form control components (title, author, ownerId, and
publisherId). The real work of performing a search is done in the BookMatches
page. All that’s left in the Home page’s Java class is the listener method used to link
the two pages together. This method, search(), is shown in listing 10.4.
Listing 10.4 The search() method of the Home class
public void search(IRequestCycle cycle)
{
BookMatches matches =
(BookMatches) cycle.getPage("BookMatches");
MasterQueryParameters parameters =
new MasterQueryParameters(getTitle(), getAuthor(),
getOwnerId(), getPublisherId());
matches.performQuery(parameters, cycle);
}
The search() method packages up the four search constraints (the title, the
author, the owner, and the publisher) as an instance of MasterQueryParameters,
then invokes the performQuery() method on the BookMatches page. MasterQueryParameters is simply a class for holding all the parameters for this search. It is easier and more extensible to pass a single object than four separate parameters.
The BookMatches page is responsible for taking it from there. It performs
the query and then displays the results.
419
This basic Tapestry approach should start looking awfully familiar by now:
Components on the page read and update properties of the page (often specified in the page’s specification), then listener methods read those same properties and act on them, possibly invoking methods on other pages within the
application or on other objects entirely.
10.4 Implementing the BookMatches page
The BookMatches page is quite a bit more complex than the Home page. It has
to run the query against the database and present the results, complete with support for paging and column sorting. It must know if the user is logged in and
support borrowing of books directly from the book listing.
The BookMatches page has four properties related to paging and sorting,
identified in table 10.3. All four of these properties are persistent.
Table 10.3
Properties used for displaying a sorted list of books
Property
Type
Description
bookQuery
org.apache.tapestry.vlib.
ejb.IBookQuery
Reference to BookQuery stateful
session bean (remote interface)
sortColumn
ejb.SortColumn
Enum for defining which column to
sort on
descending
boolean
Flag for ascending vs. descending sort
queryParameters
MasterQueryParameters
Defines constraints on query
The core of this page is the bookQuery property, which stores a reference to the
remote interface of the BookQuery stateful session bean. The BookQuery bean performs one of several possible queries against the database, merging together
data from the Book, Person, and Publisher entities.
The BookMatches page invokes the BookQuery bean’s masterQuery()
method, passing in the query parameters and the desired sort ordering (in
terms of a column to sort on and a descending flag). The BookQuery bean constructs and executes a JDBC statement for the query, and then caches the complete result in memory, returning just the number of matching books. This
result count is passed to the Browser component, which stores the count and
computes the number of pages needed to display the results. Figure 10.6 identifies these steps.
420
CHAPTER 10
Figure 10.6 The page invokes masterQuery() and then informs the Browser
component of the number of books.
The Browser component is responsible for handling pagination. It is also
responsible for communicating with the BookQuery bean in order to obtain a subset of the full results that should be displayed on the current page. We cover this
in greater detail shortly.
10.4.1 Handling paging and column sorting
So, what about re-sorting? To re-sort the list of books, the sortColumn and/or
descending properties are updated, and the query is executed again. The
BookQuery bean will then construct and execute a new JDBC statement and store
a new set of results. When the page renders, the Browser component gets a
subset of Book objects from the new set of cached results stored by the BookQuery bean.
At this point we have shown how to start a query, as well as how to handle
things when the sort ordering changes, but we don’t know how the links and
icons appear in the user interface to trigger sort order changes. We also know
that the Browser component somehow manages paging, but it isn’t clear how the
421
page navigation menu appears (figure 9.4). The answers to these questions are
in the page’s HTML template (shown in listing 10.5) and specification, and, of
course, the answers involve components, methods, and properties.
Listing 10.5 BookMatches.html: HTML template for the BookMatches page
<body jwcid="@Border" title="Matching Books"
browser="ognl:components.browser">
<table class="data">
<tr>
<th><span jwcid="sortTitle">Title</span></th>
<th><span jwcid="sortAuthor">Author</span></th>
<th><span jwcid="sortPublisher">Publisher</span></th>
<th><span jwcid="sortOwner">Owner</span></th>
<th><span jwcid="sortBorrower">Borrowed By</span></th>
<th>Borrow</th>
</tr>
<tr jwcid="browser@Browser"
query="ognl:bookQuery"
value="ognl:currentMatch"
listener="ognl:listeners.requery">
<td><a href="#" jwcid="@BookLink" book="ognl:currentMatch">
Swing Second Edition</a>
</td>
<td><span jwcid="@Insert" value="ognl:currentMatch.author">
Robinson and Vorobiev</span>
</td>
<td><span jwcid="@Insert"
value="ognl:currentMatch.publisherName">
Manning</span>
</td>
<td><a jwcid="ownerLink" href="#">Jim Moran</a>
</td>
<td><a jwcid="holderLink" href="#">Howard Lewis Ship</a>
</td>
<td class="icon"><a href="#" jwcid="@Borrow"
book="ognl:currentMatch">
<img src="images/checkout.png" alt="Borrow"
border="0" vspace="2"></a>
</td>
</tr>
</table>
b
c
</body>
</html>
d
e
422
CHAPTER 10
b
c
d
e
These components allow the columns to be sorted; they are fully declared in the
page specification.
The Browser component acts much like a Foreach component.
The BookLink component creates a link to a particular book (the ViewBook
page); the link may be bookmarked.
These two components are PersonLink components, which create “bookmarkable” links to particular persons (in the ViewPerson page).
The pieces of the puzzle start falling together here. First we have an answer to
the question of how the paging navigation controls appear. On this page, the
browser parameter of the Border component is used; this appears in the
HTML template:
<body jwcid="@Border" title="Matching Books"
browser="ognl:components.browser">
Before (in chapter 5), we bound a ValidField component to the field parameter
of a FieldLabel component; we can here bind a Browser component to the
browser parameter of the Border component. This is another powerful and useful concept: components as parameters to other components. It allows different
components (here the Border and Browser components) to cooperate with each
other in a flexible and dynamic fashion.
The Border component’s browser parameter is optional (remember, the
Home page didn’t specify a value, since there is no Browser component on the
Home page). When the browser parameter is not bound to null, the Border component includes the paging navigation controls. Later we’ll go into more detail
on exactly how the Border and Browser components interact.
Next inside the HTML template are controls for updating the table’s sort
order. These appear in the header line of the table:
<th><span
<th><span
<th><span
<th><span
<th><span
jwcid="sortTitle">Title</span></th>
jwcid="sortAuthor">Author</span></th>
jwcid="sortPublisher">Publisher</span></th>
jwcid="sortOwner">Owner</span></th>
jwcid="sortBorrower">Borrowed By</span></th>
As you can see, very little of the sorting functionality is specified in the template.
The page specification will declare the five components (one for each sortable
column) that control the sort order, so let’s look at the relevant portions of the
page specification. First, we define the two persistent properties for controlling
the sort order, giving an initial value for the sortOrder property (the descending
property will default to false):
423
<property-specification name="sortColumn"
type="org.apache.tapestry.vlib.ejb.SortColumn"
persistent="yes">
@org.apache.tapestry.vlib.ejb.SortColumn@TITLE
</property-specification>
<property-specification name="descending"
type="boolean"
persistent="yes"/>
The SortColumn class is an enumeration of the different columns that can be
sorted in any of the queries supported by the BookQuery bean. Each of the ColumnSorter components will update the page’s sortColumn property to a different
value. The first of these components, used for sorting by title, binds the page’s
sortColumn and descending properties. We use the OGNL syntax for referencing
a public static field of a class to ensure that the component and the page’s Java
class agree on names.
The definition of the sortTitle component within the page specification fills in
the details of the communication between the component and the containing page:
<component id="sortTitle" type="ColumnSorter">
<binding name="selected" expression="sortColumn"/>
<binding name="sortColumn">
@org.apache.tapestry.vlib.ejb.SortColumn@TITLE
</binding>
<binding name="descending" expression="descending"/>
<binding name="listener" expression="listeners.requery"/>
</component>
The component type, ColumnSorter, is an application-specific component, just
like the Border component. We’ll discuss its implementation shortly. The
ColumnSorter’s selected parameter is bound to the sortColumn property of the
page; this allows the sortColumn property to be updated when the ColumnSorter’s link is clicked. Likewise, the ColumnSorter will read and update the
descending property via its own descending parameter.
The sortColumn parameter of the ColumnSorter is bound to the value used to
update the selected parameter when the link is clicked (this is a usage pattern
similar to the selected and tag parameters of the Submit component, covered in
chapter 3). Each of the components declared in the BookMatches page specification will use a different value for the sortColumn parameter, resulting in the
page’s sortColumn property being updated to a different value when the corresponding link is clicked.
Updating the sortColumn and descending properties does not, of itself, cause
any change to the cached data stored within the BookQuery bean. It is necessary
424
CHAPTER 10
to force the BookQuery bean to perform a new query, using the updated sort
ordering. This is the job of a listener method, as specified using the ColumnSorter’s listener parameter. The listener is notified after the sortColumn and
descending properties have been changed. The requery() method contains the
code that runs the revised query (with the same query parameters but with a new
sort ordering) in the BookQuery bean.
The other four columns are configured in the same way, with only the sortColumn parameter changing from one component to the next. In Tapestry page
and component specifications, there’s a shortcut for handling this kind of duplication within a specification, utilizing the copy-of attribute:
<component id="sortAuthor" copy-of="sortTitle">
@org.apache.tapestry.vlib.ejb.SortColumn@AUTHOR
</binding>
</component>
<component id="sortPublisher" copy-of="sortTitle">
@org.apache.tapestry.vlib.ejb.SortColumn@PUBLISHER
</binding>
</component>
<component id="sortOwner" copy-of="sortTitle">
@org.apache.tapestry.vlib.ejb.SortColumn@OWNER
</binding>
</component>
<component id="sortBorrower" copy-of="sortTitle">
@org.apache.tapestry.vlib.ejb.SortColumn@HOLDER
</binding>
</component>
Each of the other four components simply copies the component type and
parameter bindings of the sortTitle component, then replaces the binding for
the sortColumn parameter with a different binding, constructed around a different constant value.
Consider for a moment what we’re putting together here; the page has a pair
of properties that control the sort order for the book listing, but it doesn’t know,
and doesn’t need to know, how these properties get set. At the same time, the
sorting components have an almost independent life within the page: They use
the knowledge of the current sort ordering (as expressed in that pair of properties) to decide how to render, and to decide what to do when the links they
425
themselves generate are clicked. When all of this is tied together, the end result is
columns that sort within the table—and all of this is completely reusable in other
pages that also display a list of books (even when the columns displayed are
somewhat different). That’s a lot of reuse, so let’s see how these other components are themselves used and implemented.
10.4.2 Using the Browser component
The Browser component is used, in the page’s HTML template, in much the
same way as a Foreach component. It wraps around a block of the template,
which is repeatedly rendered:
<tr jwcid="browser@Browser"
query="ognl:bookQuery"
value="ognl:currentMatch"
listener="ognl:listeners.requery">
<td><a href="#" jwcid="@BookLink" book="ognl:currentMatch">
Swing Second Edition</a>
</td>
value="ognl:currentMatch.author">
Robinson and Vorobiev</span>
</td>
value="ognl:currentMatch.publisherName">
Manning</span>
</td>
<td><a jwcid="ownerLink" href="#">Jim Moran</a>
</td>
<td><a jwcid="holderLink" href="#">Howard Lewis Ship</a>
</td>
<td class="icon"><a jwcid="@Borrow"
book="ognl:currentMatch" href="#">
<img src="images/checkout.png" alt="Borrow"
border="0" vspace="2"></a>
</td>
</tr>
The Browser is represented in the template as a <tr> element. It has three
parameters. The first, query, is an instance of IBookQuery (the BookQuery bean’s
remote interface) from which a portion of the query result set (the list of books
matching the query) can be obtained. This parameter fills a role similar to the
Foreach component’s source parameter.
The value parameter is used just like the Foreach component’s value parameter: The property bound to the value parameter is updated just before the body
of the Browser is rendered. On the BookMatches page, the currentMatch property is bound to the value parameter.
426
CHAPTER 10
The last parameter, listener, is used only if there is a RemoteException communicating with the query. This could occur if the application server has discarded the BookQuery EJB due to lack of use. Whatever the reason, the listener
parameter is used by the Browser component to request that the BookMatches
page perform the query again. The listener parameter is connected to the same
requery() method as the ColumnSorter components.
Three additional application-specific components are used on the page. The
BookLink component creates a link to the ViewBook page for a particular Book.
This link is built around the title of the book, and the URL for the link may be
bookmarked (it makes use of the external service).
The PersonLink component does an equivalent job, creating a link to the ViewPerson page for a particular person. The URL for the link may also be bookmarked.
The last component is the Borrow component, which creates a link and icon
for borrowing a book. The Borrow component contains the checks for whether
the user is logged in and whether the user is already in possession of the book.
10.4.3 Executing queries and re-queries
Although the components on the page do most of the work, they can’t quite do it
all. The BookMatches class (shown in listing 10.6) is responsible for performing
the actual query when the search form (on the Home page) is submitted or when
a change in sort ordering occurs.
Listing 10.6 BookMatches.java: Java class for the BookMatches page
package org.apache.tapestry.vlib.pages;
import java.rmi.RemoteException;
import
import
import
import
import
import
import
import
import
org.apache.tapestry.vlib.IMessageProperty;
org.apache.tapestry.vlib.VirtualLibraryEngine;
org.apache.tapestry.vlib.components.Browser;
org.apache.tapestry.vlib.ejb.IBookQuery;
org.apache.tapestry.vlib.ejb.MasterQueryParameters;
org.apache.tapestry.vlib.ejb.SortColumn;
org.apache.tapestry.vlib.ejb.SortOrdering;
public abstract class BookMatches extends BasePage
{
private Browser _browser;
public void finishLoad()
{
_browser = (Browser) getComponent("browser");
}
Stores reference to
Browser component
public abstract IBookQuery getBookQuery();
public abstract void setBookQuery(IBookQuery bookQuery);
public abstract SortColumn getSortColumn();
public abstract boolean isDescending();
public abstract MasterQueryParameters getQueryParameters();
public abstract void setQueryParameters(
MasterQueryParameters queryParameters);
public void performQuery(
MasterQueryParameters parameters,
{
setQueryParameters(parameters);
int count = executeQuery();
if (count == 0)
{
IMessageProperty page =
(IMessageProperty) cycle.getPage();
page.setMessage(getMessage("no-matches"));
return;
}
Is invoked by
Home page
_browser.initializeForResultCount(count);
cycle.activate(this);
}
public void requery(IRequestCycle cycle)
{
int count = executeQuery();
if (count != _browser.getResultCount())
Is invoked when
sort order or
direction changes
}
private int executeQuery()
{
(VirtualLibraryEngine) getEngine();
427
428
CHAPTER 10
MasterQueryParameters parameters = getQueryParameters();
SortOrdering ordering = new SortOrdering(getSortColumn(),
isDescending());
int i = 0;
while (true)
{
try
{
IBookQuery query = getBookQuery();
if (query == null)
{
query = vengine.createNewQuery();
setBookQuery(query);
}
return query.masterQuery(parameters,
ordering);
Performs
query (with
retry loop)
}
catch (RemoteException ex)
{
vengine.rmiFailure(
"Remote exception processing query.",
ex, i++);
setBookQuery(null);
}
}
}
}
Because the page must invoke methods on the Browser component in several
places, a reference to the Browser component is kept in an instance variable,
_browser. Previously we’ve said that your code should not store references to
pages or components inside persistent page properties, and this requirement
still holds. It is acceptable, as in this case, to keep a reference to a component
within the same page hierarchy, as long as it is not stored persistently. This is just
a simple optimization; the value stored in the _browser instance variable is the
exact value that would be returned by invoking getComponent("browser"). This
optimization removes the need for the method call and the typecast.
The appropriate place to create such a reference is within an implementation
of the finishLoad() method. The finishLoad() method is invoked after the page
has been created and configured, and all components for the page (specified
429
either in the template or in the page specification) have also been instantiated
and configured. This means that finishLoad() is invoked just once, and long
before other methods, such as render(), are invoked.
The performQuery() method is invoked by the Home page to start a search.
The method begins by storing the query parameters into the persistent queryParameters property. This property must be persistent because the parameters
will be needed later, if a change in sort order occurs, so that the query can be
executed with the new sort order.
The executeQuery() method executes the query and returns the result
count, the number of books that matched the search query parameters. If the
count is zero, the user is returned to the Home page with a message that there
were no matches. The user can than modify the search parameters and attempt
another search.
Normally, there will be at least one matching book. In that case, the Browser
is informed of the new result count so that it can calculate the number of pages.
Lastly, the current page, the BookMatches page, is activated as the responserendering page to render the response to the client web browser.
The requery() method, invoked as a listener method by the ColumnSorter
components, is even simpler. It invokes the executeQuery() method to perform
the query with the updated sort ordering, then informs the Browser if the result
count has changed. When the result count changes, the current page is always
reset to 1, so it’s best not to invoke the initializeForPageCount() method on the
Browser instance unless the result count actually does change.
The leaves just the executeQuery() method itself. This method first assembles
the parameters to the BookQuery beans’ masterQuery() method: the MasterQueryParameters instance and a SortOrdering instance (built from the sortColumn
and descending page properties).
Next, a reference to the BookQuery bean’s remote interface is obtained, if
needed. Since many pages utilize the BookQuery EJB, the code to find and create
an instance is centralized in the VirtualLibraryEngine class. The method then
executes the masterQuery() method on the bean and returns the result, the count
of matching books.
The only complication is the retry loop, used to ride out any exceptions
related to invoking the remote method on the BookQuery bean.4 In the event
4
In the deployed application, the presentation and application layers are in the same JVM, so RemoteExceptions are rare. This retry code is most useful when the layers are separate, which can frequently
occur when testing.
430
CHAPTER 10
of a RemoteException , the VirtualLibraryEngine ’s rmiFailure() method is
invoked. The attempt index (the variable i) is passed in. On the first attempt,
the error is logged, and rmiFailure() returns, to allow a retry. On the second
attempt, an actual failure occurs, and the user will see an exception report
page. This approach to RemoteException recovery recurs throughout the Virtual Library application.
Also, on a failure, the BookQuery bean reference is discarded, forcing a new
BookQuery bean to be obtained from the engine. This addresses the most likely
cause of a RemoteException: that the BookQuery bean, a stateful session bean, has
been discarded by the application server due to lack of use. Invoking methods
on a discarded remote reference results in a RemoteException that can be rectified by creating a new instance of the EJB.
Here robustness and usability go hand in hand. The user should not need to
know that the application uses stateful session beans, or that such beans may be
discarded by the application server even as the rest of the application continues
to run. Instead, the user will get the results they expect: Clicking on the column
titles will re-sort the columns. Building this level of robustness into a Tapestry
application takes care and some foresight but is quite manageable. This is one
area where the framework’s advantages shine: Because all the Java code is inside
real Java classes (as opposed to scriptlets inside a JSP), it is possible to write
robust, well-organized code. As we continue looking at the other components
and objects inside the Virtual Library, you’ll see how these individual islands of
code work together to form a robust whole.
10.5 Implementing the Browser component
The Browser component is an integral part of several pages in the application:
BookMatches, MyLibrary, BorrowedBooks, and ViewPerson. Each of these pages
works with its own instance of the BookQuery bean to present a list of books. Each
one supports paging access to the overall list of books, and each supports sorting
by column.
10.5.1 Specifying Browser’s parameters
The Browser component is half of the paging navigation equation; it is the half
that actively communicates with the BookQuery bean, as a part of the page that
contains it, and renders a subset of the query results cached by the bean.
431
Listing 10.7 Browser.jwc: specification for the Browser component
allow-informal-parameters="yes"
class="org.apache.tapestry.vlib.components.Browser">
<parameter name="query"
required="yes"
type="org.apache.tapestry.vlib.ejb.IBookQuery"
direction="auto"/>
<parameter name="value"
type="java.lang.Object"
required="yes"
direction="auto"/>
<parameter name="element"
direction="in"/>
<parameter name="listener"
type="org.apache.tapestry.IActionListener"
required="yes"
direction="auto"/>
<property-specification name="resultCount"
type="int"
persistent="yes"/>
<property-specification name="currentPage
type="int"
persistent="yes"/>
name="pageCount" type="int"
persistent="yes"/>
<property-specification name="pageResults"
type="java.lang.Object[]"/>
The component specification for the Browser component, shown in listing 10.7,
declares the three parameters we’ve previously discussed, as well as an additional
parameter, element, which controls the HTML element that will be used when
rendering the component (the default for which is tr). The query, value, and
432
CHAPTER 10
listener parameters all use the auto value for their direction attribute, for rea-
sons we’ll cover shortly.
In addition, several properties are declared. The resultCount, pageCount, and
currentPage properties are persistent. These properties store the total number of
books in the query, the number of pages required to display all the books, and
the current page (numbered from 1) within the range of pages.
The final property, pageResults , is not persistent. It is an array of Book
objects, the Books to display on the current page. It may not immediately be obvious why this is necessary—it’s related to error recovery mixed with persistent
page properties.
10.5.2 Getting results from the BookQuery bean
The Browser component must invoke the get() method of the BookQuery bean
to obtain the results for the current page, but as we’ve discussed, the request
may fail for a number of reasons. When the request fails, the Browser notifies
its listener, which will in all likelihood create a new BookQuery bean instance
and update the persistent property used to store the reference to the bean’s
remote interface.
There’s the problem: Updates to persistent properties are not allowed once
the rendering of a response has started (this is covered in detail in chapter 7).
Therefore, the access to the BookQuery bean must occur before the page renders,
since it is possible that the persistent property storing the reference to the bean
will change.
This is accomplished by having the Browser component implement the PageRenderListener interface. The pageBeginRender() method is invoked before the
page starts to render, when it is still allowable to change persistent properties. The
page results, obtained from the BookQuery bean, are stored in the pageResults
property at this point, as shown in figure 10.7. Later, when the Browser component
renders, the results will be ready and waiting in that property.
The pageBeginRender() method includes the same kind of retry loop used by
the BookMatches page. In this case, the recovery from a RemoteException accessing the BookQuery bean will invoke the listener method to perform a new query.
This will result in a second execution of the loop, and getQuery() will return the
new BookQuery bean instance. The results obtained from the BookQuery bean are
then stored in the pageResults property.
This method is invoked outside the normal rendering of the component.
Therefore, the connected parameter properties that are referenced in the
method (query and listener) must specify a direction attribute of auto, not in.
433
Figure 10.7
Before the page
renders, the Browser
component gets and
stores the books to be
displayed on the page.
10.5.3 Rendering the Browser component
All this work in the pageBeginRender() method means that the renderComponent()
method (provided in listing 10.8) is very simple. All it has to do is obtain the
array of books as stored in the pageResults property and repeatedly render its
body, updating its value parameter for each Book object provided by the BookQuery bean.
Listing 10.8 The renderComponent() method of the Browser class
{
Object[] books = getPageResults();
int count = Tapestry.size(books);
String element = getElement();
{
setValue(books[i]);
if (element != null)
{
writer.begin(element);
}
renderBody(writer, cycle);
434
CHAPTER 10
if (element != null)
writer.end();
}
}
Because the value parameter’s direction attribute is auto, each time the property is updated (using the setValue() method), the bound property of the page
(the currentMatch property for the BookMatches page) will be updated. The
other components on the page read the currentMatch property (or properties
nested within it) when rendering.
That covers how the query is created, accessed, and rendered. The next section delves into how the page’s sortColumn and descending properties are
updated, which is encapsulated inside the ColumnSorter component.
10.6 Implementing the ColumnSorter component
Continuing our tour of the BookMatches page, we’ll next visit the ColumnSorter
component. This component is responsible for making the column titles on the
page clickable links. Clicking a link will sort the selected column in ascending
order—unless the column was already the selected column, in which case clicking a link will toggle between ascending and descending order.
In addition, the selected column gets an icon next to the title that indicates
whether the column is sorted in ascending order or descending order. The title
text comes from the body of the ColumnSorter component.
The ColumnSorter component leverages Tapestry’s component-level request
dispatching; it’s a component that contains components, and these internal
components have interactions with the end user that the containing page is completely unaware of!
10.6.1 Creating the ColumnSorter HTML template
The template for the ColumnSorter component (shown in listing 10.9) is primarily an <a> link element wrapped around a RenderBody component. Beyond
that, the template includes additional elements that display the sort icon.
Listing 10.9 ColumnSorter.html: HTML template for the ColumnSorter component
<a jwcid="link"><span jwcid="@RenderBody"/></a>
<span jwcid="@Conditional" condition="ognl:sortColumn == selected">
<a jwcid="iconLink"><img jwcid="rollover"
435
alt="Resort Column"/></a>
</span>
If the selected SortColumn for the page matches the SortColumn for this ColumnSorter component, then the second link and the matching icon are also displayed. Using two separate links is necessary only for presentation reasons. If a
single link enclosed the RenderBody component, the Conditional component,
and the icon, then there would be a visible, underlined space between the page
title and the icon.
The specification for the ColumnSorter component defines the two links and
introduces a new type of component, Rollover. Rollover is another component
provided with the framework. It is a special version of the Image component that
must be enclosed by some kind of Tapestry link component (here it is the
DirectLink component, but it could be PageLink or ExternalLink). The Rollover
component performs mouse rollover effects; that is, the image changes when the
user moves the cursor over the image, and changes back when the cursor is
moved off it. This provides additional feedback to the user about which elements
on the page may be clicked. Rollover components are used throughout the Virtual Library application.
10.6.2 Specifying ColumnSorter parameters
The specification for the component (in listing 10.10) declares all the parameters for the ColumnSorter, as well as the assets used for the sort icon.
Listing 10.10 ColumnSorter.jwc: specification for the ColumnSorter component
class="org.apache.tapestry.vlib.components.ColumnSorter"
allow-body="yes" allow-informal-parameters="no">
<parameter name="selected"
required="yes"
direction="auto"/>
<parameter name="descending"
type="boolean"
required="yes"
436
CHAPTER 10
direction="auto"/>
<parameter name="sortColumn"
required="yes"
direction="auto"/>
<parameter name="listener"
type="org.apache.tapestry.IActionListener"
required="yes"
direction="auto"/>
<component id="link" type="DirectLink">
<binding name="listener" expression="listeners.handleClick"/>
</component>
<component id="iconLink" copy-of="link"/>
<component id="rollover" type="Rollover">
descending ? assets.down : assets.up
</binding>
<binding name="focus">
descending ? assets.down_h : assets.up_h
</binding>
</component>
<context-asset
<context-asset
<context-asset
<context-asset
Copies link
as iconLink
Defines a
Rollover for
the icon
name="up" path="images/sort-up.png"/>
name="up_h" path="images/sort-up_h.png"/>
name="down" path="images/sort-down.png"/>
name="down_h" path="images/sort-down_h.png"/>
A Rollover component has four parameters for specifying image assets to be
used with the component. The three common parameters are image, focus, and
disabled. The image parameter specifies the image the application should normally display and is the only required parameter. The focus parameter specifies
the image to display when the mouse is moved over the image.5 The disabled
parameter specifies the image the application should display when the link
enclosing the Rollover component is disabled (in which case the Rollover component does not react to the mouse). A fourth parameter, blur, specifies the
5
This is another example of a less-than-ideal name getting wedged into the framework. Changing the
parameter name to something like “mouseOver” would make sense, but that would break too many
existing components and applications.
437
image to use when the mouse is moved out from over the image (there’s rarely a
reason for this to not be the same as the image parameter, so it’s usually omitted).
The upshot of this is that it is easy to create clickable image buttons that have
dynamic rollover effects. It is simply a matter of defining the necessary assets.
The convention used in the Virtual Library is to add an _h suffix to the name of
an asset to indicate a highlighted version of an image—one appropriate as a
mouse-over image (shown here as the focus parameter).
The specification declares four image assets: two versions of the up arrow and
two of the down arrow. The bindings for the Rollover component dynamically
determine which assets to use for the image and focus parameters based on
whether the sort order is ascending or descending. Once again, the use of OGNL
expressions allows for a succinct, readable solution to the problem of selecting
the correct assets.
10.6.3 Responding to the user
All that’s left is a listener method to handle things when either link is clicked. The
same listener method, handleClick() (in listing 10.11), is invoked by both links.
Listing 10.11 The handleClick() method of the ColumnSorter class
public void handleClick(IRequestCycle cycle)
{
SortColumn selected = getSelected();
SortColumn sortColumn = getSortColumn();
if (selected != sortColumn)
{
setSelected(sortColumn);
setDescending(false);
}
else
{
boolean current = isDescending();
setDescending(!current);
}
IActionListener listener = getListener();
if (listener == null)
throw Tapestry.createRequiredParameterException(this,
"listener");
listener.actionTriggered(this, cycle);
}
438
CHAPTER 10
The handleClick() method first checks to see if the currently selected SortColumn
matches the sortColumn parameter for this ColumnSorter component. If so, the
descending parameter is inverted. If it is a non-match, meaning the user has
selected as new column as the column to sort, then the sortColumn parameter is
updated to the correct value, and the descending parameter is set to false.
In either case, the listener method is invoked. As you saw in the BookMatches
page, the listener method’s job is to force the query to be executed again with
the updated sort ordering.
The final piece of the Browser puzzle is the browser navigation controls
(which are shown in figure 9.4). Presenting these controls on a page demands
cooperation between the Browser component and the Border component.
10.7 Implementing the Border component
At this point, we have enough of the pieces of the application to discuss how to
construct the Border component. A considerable amount of the look and feel of
the application, as well as its functionality, is tied up in this single component.
This Border component is responsible for displaying the proper page title
(across the top of the page), the main navigation menu on the left (which adjusts
to reflect the page and the user), and an optional page navigation menu for
pages that use a Browser component (shown in figure 9.4). In fact, the Border
component is considerably larger and more complicated than any of the pages
that use it.
Most of the Border component’s behavior is concerned with the navigation
menu. The menu is dynamic in several ways:
■
All the images in the menu are rollover buttons, highlighting as the mouse
is moved over them.
■
The current page is displayed with an alternate image incorporating a
small arrow.
■
The My Library menu items appear only once the user has logged in.
■
Virtual Library application administrators will see three additional
menu items.
■
The final option in the menu may be Login or Logout depending on
whether the user has, in fact, authenticated (logged in or registered).
Despite this highly dynamic nature at runtime, the implementation of the Border component is extremely straightforward. Each menu item consists of a link
439
Figure 10.8
Here are four versions
of the Search button.
and an image in the HTML template and the specification, plus additional assets
for the various button images. For example, the first menu item is the Search
link. This link is represented by four images, shown in figure 10.8. One pair of
images (either the normal and highlighted, or the selected and selected highlighted) will be used. On a Search-related page (Home, BookMatches, ViewBook, and ViewPerson), the second pair of images will be used. The arrow in the
image indicates that the user is in the “Search” section of the application. When
other pages render, such as MyLibrary or Login, the first pair of images is used.
In the Border component’s HTML template, menu item appears as some form
of link component and a Rollover component (both are declared components):
<a href="#" jwcid="search"><img
jwcid="searchRollover" src="images/nav/nav_1x1.png"
width="178" height="29" border="0" alt="Search"/></a>
The Border component’s specification includes these two components, along
with the four assets for the Search button image:
<component id="search" type="PageLink">
<static-binding name="page" value="Home"/>
</component>
<component id="searchRollover" type="Rollover">
<binding name="image" expression="searchImage"/>
<binding name="focus" expression="searchRolloverImage"/>
</component>
<context-asset name="search" path="images/nav/nav_1x1.png"/>
<context-asset name="search_h" path="images/nav-h/nav_1x1.png"/>
<context-asset name="search_s"
path="images/nav-selected/nav_1x1.png"/>
<context-asset name="search_h_s"
path="images/nav-selected-h/nav_1x1.png"/>
To simplify things, the names of the assets use an _h suffix to indicate a highlighted image and an _s suffix for the selected image. The Rollover component
does not reference the assets directly; instead, it references two properties of the
Border component: searchImage and searchRolloverImage. These properties are
440
CHAPTER 10
computed when the Border class, the Java class for the Border component, is
instantiated, inside the finishLoad() method. Listing 10.12 includes a portion
of the Java code for the Border class that illustrates this process.
Listing 10.12 Border.java (partial): Java class for the Border component
private
private
private
private
String
IAsset
IAsset
IAsset
_pageType;
_titleImage;
_searchImage;
_searchRolloverImage;
public void finishLoad()
{
IPage page = getPage();
String pageName = page.getPageName();
_pageType = page.getSpecification().
getProperty("page-type");
if (_pageType == null)
_pageType = pageName;
Retrieves page-type property
from page’s specification
Uses page’s name as
default for page type
_titleImage = getAsset("title_" + pageName);
if (_titleImage == null)
_titleImage = getAsset("title_" + _pageType);
if (_titleImage == null)
throw new ApplicationRuntimeException(
"Cannot find title image for " + pageName +
" or " + _pageType + ".",
this);
_searchImage = selectImage("Search", "search");
_searchRolloverImage = selectImage("Search", "search_h");
. . .
}
private IAsset selectImage(String type, String baseName)
{
String key = _pageType.equals(type)
? baseName + "_s"
: baseName;
Finds normal
or selected
asset
return getAsset(key);
}
public IAsset getTitleImage()
{
return _titleImage;
}
441
Provides read-only
property referenced
in template
public IAsset getSearchImage()
{
return _searchImage;
}
public IAsset getSearchRolloverImage()
{
return _searchRolloverImage;
}
The finishLoad() method is responsible for adapting the Border component to
the page that contains it—this method precalculates a number of properties that
are referenced in the component’s template and specification. Each page
instance will contain its own copy of the Border component—the Border component on the Home page is a completely separate instance from the Border component on the MatchingBooks page, for example. Because Tapestry components
are stateful objects, we can precalculate and store the information that will be
needed later (when the component renders), which is much more efficient and
straightforward than having to calculate all that information on the fly.
The finishLoad() method starts by determining the page type for the current page. If the page’s specification has a <property> for name page-type, then
that is used as the page type. Otherwise, the page’s name is the page type. In
most cases, the page type matches the page name, but there are a few rule breakers, such as the Home and BookMatches pages (which define their page-type
as Search).
Next, finishLoad() determines the title image to use for the page. If there’s
an image asset whose name matches the name of the page, then that image is
used. Otherwise, the image matching the page type is used.
After that, the selectImage() method is used to find the correct image asset
to use for the normal and highlighted versions of the Search button. The selectImage() method appends an _s suffix to the base asset name when the current
page type matches the type parameter. In the actual class, this continues for each
of the many dynamic images used in the navigation menu.
442
CHAPTER 10
10.7.1 Handling user login
The Border component includes a Login menu item, which allows the user to
log into the system (possibly registering on the fly) and, after authenticating,
return to the current page. This is useful on pages such as BookMatches, ViewBook, and ViewPerson, since after logging in, the user returns to the same page
and is able to borrow a book.
This is accomplished using a DirectLink component and a listener method,
login(). The source for this method is shown in listing 10.13.
Listing 10.13 The login() method of the Border class
public void login(IRequestCycle cycle)
{
Visit visit = (Visit) cycle.getEngine().getVisit();
if (visit != null && visit.isUserLoggedIn())
return;
ICallback callback = new PageCallback(getPage().getPageName());
Login loginPage = (Login) cycle.getPage("Login");
loginPage.setCallback(callback);
cycle.activate(loginPage);
}
The login() listener method checks to see if the user is already logged in. If so,
the method returns, and the active page is redisplayed. If the user has not
already logged in, the Login page is invoked to allow the user to log in.
The interesting part is how we set up the Login page returns to return to the
current page after the user login process is complete. The Login page has a
property, callback, of the type ICallback. ICallback is an interface provided
with the framework, which exists for just this purpose: to define in an abstract
way how to continue after some form of interruption. In this case, the interruption is logging into the application. The interface and a base set of implementations exist specifically to handle common use cases such as logging in,
registering, and entering optional addresses into a form. Here, that interruption
is authenticating the user, either by logging in or by registering.
The implementation used, PageCallback , will activate the named page
(invoking the IRequestCycle object’s activate() method), which is what we want
in this situation. As you’ll see when we cover the Login page in detail, the callback returns users to where they started once they are authenticated.
443
10.7.2 Linking to MyLibrary
The MyLibrary page displays a list of books owned by the logged-in user. Like
the BookMatches page, it must construct and manage a reference to a BookQuery
bean. Unlike the Home page, which can be rendered at any time, the MyLibrary
page must be initialized before it can render.
Two other pages in the application are similar: BorrowedBooks and EditProfile. Each has an initialization stage that must occur before the page can be
rendered the first time. For the Virtual Library, an interface, IActivate, exists for
these three pages. This allows all three pages to be treated similarly by the Border component.6
At the same time, it is necessary to log the user in, on the fly, before rendering
the MyLibrary page. The link to the MyLibrary page is a DirectLink, and the listener method is viewMyLibrary(). This listener method simply invokes another
method, activate(), to do the work (the Border class includes other listener
methods that invoke activate() with different parameters, supporting several
other menu items). Both of these methods are shown in listing 10.14.
Listing 10.14 The viewMyLibrary() and activate() methods of the Border class
public void viewMyLibrary(IRequestCycle cycle)
{
activate("MyLibrary", cycle);
}
private void activate(String pageName, IRequestCycle cycle)
{
IActivate page = (IActivate) cycle.getPage(pageName);
page.validate(cycle);
page.activate(cycle);
}
The viewMyLibrary() and activate() methods obtain the MyLibrary page and
invoke its validate() method. If validate() does not throw an exception, then
the page’s activate() method (as defined by the IActivate interface) is invoked.
6
Another option, which is completely reasonable, would be to create a new type of service for activating
such pages. The service could be similar to the page service but also cast the page to IActivate and
invoke the activate() method. If we were to pursue such an approach, we would also create an
ActivateLink component.
444
CHAPTER 10
The MyLibrary class includes an implementation of the activate() method,
shown in listing 10.15.
Listing 10.15 The activate() method of the MyLibrary class
public void activate(IRequestCycle cycle)
{
runQuery();
}
This purpose of this implementation of the activate() method is to execute the
query of books owned by the logged-in user (that’s primarily what’s shown on
the MyLibrary page: a listing of books the logged-in user owns) and then set the
MyLibrary page as the active page (so that it can render the response). Like
the BookMatches page, the MyLibrary page uses its own instance of the BookQuery bean and has its own set of Browser and ColumnSorter components.
The final piece of the puzzle is to force the user to be logged in before activating
the MyLibrary page. This comes from ActivatePage, the super class of MyLibrary.
ActivatePage is a base implementation of the IActivate interface. ActivatePage
also implements the PageValidateListener interface; it includes an implementation of the pageValidate() method that checks whether the user is logged in and,
if not, uses the Login page to get the user logged in. Listing 10.16 provides the
source code for this method.
Listing 10.16 The pageValidate() method of the ActivatePage class
public void pageValidate(PageEvent event)
{
return;
Login login = (Login) getRequestCycle().getPage("Login");
login.setCallback(new ActivateCallback(this));
throw new PageRedirectException(login);
}
445
Throwing a PageRedirectException forces Tapestry to activate the specified page
and use it to render the response, regardless of what the active engine service
would normally do. The ActivateCallback class implements ICallback and is
used to invoke activate() on a page that implements the IActivate interface.
This combination—the listener method in the Border component and the
methods in the MyLibrary and ActivatePage classes—ensures that the MyLibrary
page will render only after the user has successfully logged in and that the
MyLibrary page will be allowed to initialize properly before being rendered.
The same protections also apply to the other (similar) pages, BorrowedBooks
and EditProfile.
Before users can access these protected pages, the application must provide a
way to authenticate them.
10.8 Authenticating the user
Authentication of the user involves
■
Collecting the user’s email address and password and validating them
against the database
■
Collecting registration information and registering the user in the database
or
After login or registration, the process returns the user to the correct page.
As you’ve seen with the MyLibrary page, the pageValidate() method is
responsible for both directing the user to the Login page and telling the Login
page how to continue after the user is authenticated. The base class Protected
also implements the pageValidate() method (shown in listing 10.17); this
method is used by pages such as the NewBook page (whose Java class extends
from Protected) that are only available to authenticated users but do not require
the initialization stage mandated by the IActivate interface.
The Protected class’s implementation of the pageValidate() method is similar to
the implementation of pageValidate() in the ActivatePage class, except that the callback is an instance of PageCallback, not ActivateCallback. Using the PageCallback
class is appropriate for pages that don’t implement the IActivate interface.
Listing 10.17 The pageValidate() method of the Protected class
public void pageValidate(PageEvent event)
{
446
CHAPTER 10
return;
Login login = (Login) getRequestCycle().getPage("Login");
login.setCallback(new PageCallback(this));
throw new PageRedirectException(login);
}
Here again, the Visit object is accessed to see if the user has already authenticated. If not, the Login page is rendered, with a callback to return the user to the
current page.
10.8.1 Remembering the user
When users launch the application at a later date, the correct email address will
be filled out for them and they’ll just have to enter a password. This is a natural
use for an HTTP cookie. Although Tapestry applications can be coded without
referencing the Servlet API, you are not prevented from doing so; the Login
page uses portions of the Servlet API to read and set an HTTP cookie that
records the identity of the user for later sessions.
The Login page implements the PageRenderListener interface so that it may be
notified when it is about to render. The pageBeginRender() method checks to see if
the email address is null and supplies a default value from the HTTP cookie.
The initialization code in listing 10.18 is balanced with code that records the
cookie after a successful login, which is shown in listing 10.19 in the next section.
Listing 10.18 The pageBeginRender() method of the Login class
private static final String COOKIE_NAME =
"org.apache.tapestry.vlib.Login.email";
{
if (getEmail() == null)
setEmail(
getRequestCycle().getRequestContext().
getCookieValue(COOKIE_NAME));
}
447
10.8.2 Clearing the password field
When a form within a page is submitted and the page redisplayed, normally the
TextField and ValidField components within the page redisplay the input that
was supplied by the end user. For TextField, the input is stored in a page property when the form is submitted and then read from the same property when the
component renders. ValidField does the same, with the extra step of recording
user input exactly as supplied by the user. This input will be used if the user
makes an error (for instance, entering a string value into a numeric field).
This is not desirable for the password field. On a login form, if the user enters
an incorrect password, the password field should be cleared, forcing the user to
reenter the correct password from scratch. Because a ValidField component is
used to handle the password, an extra step is needed beyond simply setting the
page’s password property to null.
Listing 10.19 The attemptLogin() and loginUser() methods of the Login class
public void attemptLogin(IRequestCycle cycle)
{
String password = getPassword();
setPassword(null);
getValidationDelegate();
delegate.setFormComponent(
(IFormComponent) getComponent("inputPassword"));
delegate.recordFieldInputValue(null);
Clears
password field
return;
int i = 0;
while (true)
{
try
{
IOperations operations =
vengine.getOperations();
Person person =
operations.login(getEmail(), password);
Performs
login using
Operations
bean
448
CHAPTER 10
loginUser(person, cycle);
break;
}
catch (LoginException ex)
{
String fieldName =
ex.isPasswordError()
? "inputPassword"
: "inputEmail";
Completes
login
setErrorField(fieldName, ex.getMessage());
return;
}
{
vengine.rmiFailure(
"Remote exception validating user.",
ex, i++);
}
}
}
public void loginUser(Person person, IRequestCycle cycle)
throws RemoteException
{
String email = person.getEmail();
visit.setUser(person);
ICallback callback = getCallback();
if (callback == null)
cycle.activate("Home");
else
callback.performCallback(cycle);
Stores logged-in
user in Visit
Returns to
previously
active page
IEngine engine = getEngine();
Cookie cookie = new Cookie(COOKIE_NAME, email);
cookie.setPath(engine.getServletPath());
cookie.setMaxAge(ONE_WEEK);
Stores cookie
in client’s
browser
cycle.getRequestContext().addCookie(cookie);
engine.forgetPage(getPageName());
}
Discards persistent
properties
449
The validation delegate stores the input value, the exact value supplied by the
user, for each ValidField. For the password field to be rendered as empty, it is necessary to override the value supplied by the user with null. This can be seen at the
beginning of the attemptLogin() method (listing 10.19). Overriding the validation delegate’s stored value and setting the page’s password property to null
ensures that if the page is rendered again (for example, if the password is not
valid or if the email address is not known), the password field will be empty.
10.8.3 Invoking the login operation
The remainder of the login form’s listener obtains a reference to the Operations
bean and invokes the login() method on it. This method may throw a LoginException if the email address is unknown or the password is invalid (as a convenience for the presentation layer, this exception includes a flag indicating
whether the error should be attributed to the password).
After the user is successfully authenticated, the listener method invokes the
loginUser() method to finalize the login. This includes using the callback property to return to the original page where the user clicked the Login (or My
Library) menu item.
After the callback is invoked, the Login page calls the forgetPage() method
on the engine. This method discards all persistent properties for the page—in
this case, the single persistent property, callback . Using the forgetPage()
method is the correct way to fully eliminate persistent state that is no longer necessary. For this application, once the callback is utilized, there’s no longer any
need to store it.
The loginUser() method is public because it is also used by the Register
page. After a user registers, the loginUser() method is used to store the user’s
identity in the Visit object and invoke the callback, just as if the user already
existed and had simply logged in.
We’ve set the stage now for protecting pages by restricting them to logged-in
users, and we’ve given users the ability to log in or register on the fly. Now let’s
look at a way to store links to selected pages and data as web browser bookmarks.
10.9 Creating bookmarkable links
Links to both books and persons appear throughout the Virtual Library application. Every book and person listed on the BookMatches, BorrowedBooks, ViewPerson, and ViewBook pages appears as a link to either the ViewBook or ViewPerson
page (as appropriate).
450
CHAPTER 10
These links are “bookmarkable,” meaning that they may be saved in a user’s
bookmark list indefinitely. Bookmarkable links can be created using the external
service, and they are expected to be retrieved at a later date—unlike links created with the direct service (using the DirectLink component), which are
expected to be used only within a single application session.
Generally, external service links are created using the ExternalLink component. The ExternalLink component will only generate the link itself; by wrapping the ExternalLink inside an application-specific component, it is possible to
generate the text inside the link as well.
10.9.1 Creating the BookLink component
The BookLink component creates a link to the ViewBook page. The component
has a single parameter, book, of the type Book. The component’s template inserts
the Book’s title inside a link. It also checks to see whether the book is new; new
books are marked with an additional icon. The HTML template for this component is shown in listing 10.20.
Listing 10.20 BookLink.html: HTML template for the BookLink component
<a href="#" jwcid="link"><span
jwcid="@Insert" value="ognl:book.title"/></a>
condition="ognl:newlyAdded"><img src="images/new.png"
width="27" height="12" border="0" alt="New"></span>
The decision on whether a book is “new” is complicated by the fact that the user
may not yet be logged in. Normally, only books added since the user’s last visit
are considered new. If the user hasn’t logged in, then anything added in the previous week is considered new. This logic is encapsulated in the isNewlyAdded()
method (listing 10.21) of the BookLink class.
Listing 10.21 The isNewlyAdded() method of the BookLink class
private static final long ONE_WEEK_MILLIS =
1000l * 60l * 60l * 24l * 7l;
public boolean isNewlyAdded()
{
IEngine engine = getPage().getEngine();
Visit visit = (Visit) engine.getVisit();
Timestamp lastAccess = null;
451
if (visit != null)
lastAccess = visit.getLastAccess();
Book book = getBook();
Timestamp dateAdded = book.getDateAdded();
if (dateAdded == null)
return false;
if (lastAccess == null)
{
long now = System.currentTimeMillis();
return (now - dateAdded.getTime()) <= ONE_WEEK_MILLIS;
}
return lastAccess.compareTo(dateAdded) <= 0;
}
The link component passes the Book’s ID as a service parameter. This will be
enough for the ViewBook page to read the Book object when the page is triggered by the URL. The link component is declared in the BookLink’s component specification:
<component id="link" type="ExternalLink">
<static-binding name="page" value="ViewBook"/>
<binding name="parameters" expression="book.id"/>
</component>
That covers how to create the link and the text within the link. What happens
when the link is clicked by the user? That’s the domain of the ViewBook page.
10.9.2 Displaying the Book on the ViewBook page
The ViewBook page is the target of the ExternalLink component contained by
the BookLink component. As an external page (ViewBook implements the
IExternalPage interface, making it compatible with the external service), it
implements activateExternalPage(), the method invoked by the external service. Listing 10.22 shows the source code for this method.
Listing 10.22 The activateExternalPage() method of the ViewBook class
{
452
CHAPTER 10
Integer bookId = (Integer) parameters[0];
setBookId(bookId);
}
This method’s sole responsibility is to extract the ID of the Book entity to view
from the service parameters and store the value into the bookId page property.
When the page renders, the book will be read as needed, using the standard
approaches you’ve seen in prior Virtual Library pages. Listing 10.23 contains the
source code for the pageBeginRender() and readBook() methods.
Listing 10.23 The pageBeginRender() and readBook() methods of the ViewBook class
{
if (getBook() == null)
readBook();
}
private void readBook()
{
Integer bookId = getBookId();
int i = 0;
while (true)
{
IOperations bean = vengine.getOperations();
try
{
setBook(bean.getBook(bookId));
return;
}
catch (FinderException ex)
{
vengine.presentError("Book not found in database.",
getRequestCycle());
return;
}
{
vengine.rmiFailure(
"Remote exception obtaining information for book #" +
bookId + ".",
453
ex,
i++);
}
}
}
What happens if the user reaches the ViewBook page before logging in, and then
clicks the Login menu item? The user will be authenticated using the Login page
(and perhaps Register as well) and then will be returned to the ViewBook page using
a PageCallback object. At this point, the page will be rendered, with the Borrow
component enabled.
What’s missing in this scenario is which book should be displayed. The PageCallback stores only the name of the page to render the response; the activateExternalPage() method will not be invoked. For this reason, the bookId property
is made persistent. This ensures that when the page is rendered again after user
authentication, the correct Book object is read and displayed.
In the interests of efficiency, and in order to minimize server-side state,
only the bookId property, not the entire Book object (stored in the book property), is made persistent. The book property can always be restored from the
database, as long as the bookId is known. This follows the general guideline
of storing the minimum persistent state and recovering any additional state
as needed.
10.9.3 Creating the PersonLink component
The PersonLink component is slightly more complex than the BookLink component. Instead of passing in a Person object, a Person’s name and ID are passed in
as two separate parameters. This is because the Person object is usually not available; instead, the Book object includes properties for a book’s owner (name and
ID) and holder (name and ID). This reflects that the Book object, returned by the
BookQuery bean, is a flattening of information from the Book, Person, and Publisher entities.
In addition, there are places where a PersonLink should be omitted. For
example, on the BookMatches page, if the holder matches the owner, the holder
is omitted. This is represented as a third parameter, omit.
The HTML template for the PersonLink component, shown in listing 10.24,
makes use of these three parameters.
454
CHAPTER 10
Listing 10.24 PersonLink.html: HTML template for the PersonLink component
<span jwcid="@Conditional" condition="ognl:! omit">
<a jwcid="link"><span jwcid="@Insert" value="ognl:name">
Joe User</span></a>
</span>
<span jwcid="@Conditional" condition="ognl:omit">
 
</span>
Because the PersonLink component is often used as part of a table (that is, inside
a <td> element), it is useful to replace the link with a nonbreaking space (the
  entity) so that the browser renders the cell containing the PersonLink
properly. Different browsers display empty <td> elements differently—often the
background color or cell borders are omitted for empty table cells. Putting a
nonbreaking space inside the <td> ensures that the element is never empty.
The specification for the PersonLink component declares all three parameters. The id parameter is a connected parameter (its direction attribute has the
value in), but the id parameter may not be connected normally; the IComponent
interface already defines and uses a property named id (it stores the component’s ID, assigned by the user or by the framework). Therefore, the id parameter is connected to a different property, personId:
<parameter name="id"
type="java.lang.Integer"
required="yes"
direction="in"
property-name="personId"/>
The link component can then reference the id parameter via its connected property, personId:
<component id="link" type="ExternalLink">
<static-binding name="page" value="ViewPerson"/>
<binding name="parameters" expression="personId"/>
</component>
The other two parameters for the PersonLink component are the name parameter (the name to be displayed for the user) and the omit parameter (which displays can be used to not display the name or link). The link created references
the ViewPerson page.
455
10.9.4 Displaying the Person
The ViewPerson page displays the properties of the Person object, but it also displays the list of books owned by the person. This means that ViewPerson must
have the same kind of initialization stage that the MyLibrary page has. It must also
have additional components, properties, and Java code to allow the user to control sort order. Because the ViewPerson page is triggered using the external service, it can perform this initialization inside its activateExternalPage() method
(shown in listing 10.25) rather than implementing the IActivate interface (as
the MyLibrary page’s Java class does).
Listing 10.25 The activateExternalPage() method of the ViewPerson class
{
Integer personId = (Integer) parameters[0];
setPersonId(personId);
int count = runQuery();
}
The method extracts the id of the Person that should be displayed from the service parameters and stores it in the personId property (which, like the bookId
property of the ViewBook page, is a persistent property). It then invokes runQuery() to initialize the query that contains the list of books owned by the Person.
The runQuery() method is shown in listing 10.26.
Listing 10.26 The runQuery() method of the ViewPerson class
private int runQuery()
{
Integer personId = getPersonId();
SortOrdering ordering = new SortOrdering(getSortColumn(),
isDescending());
int i = 0;
while (true)
{
IBookQuery query = getQuery();
456
CHAPTER 10
if (query == null)
{
query = vengine.createNewQuery();
setQuery(query);
}
try
{
return query.ownerQuery(personId, ordering);
}
{
vengine.rmiFailure("Remote exception for owner query.",
ex, i++);
setQuery(null);
}
}
}
This code resembles the code you’ve already seen in the BookMatches and
MyLibrary pages. The only significant difference involves which method of the
BookQuery bean is invoked: in this case, it is the ownerQuery() method. Again
echoing the ViewBook page, the ViewPerson page initializes the person property
using the pageBeginRender() method (shown in listing 10.27).
Listing 10.27 The pageBeginRender() method of the ViewPerson class
{
Person person = getPerson();
if (person == null)
{
person = vengine.readPerson(getPersonId());
setPerson(person);
}
}
Editing a Book
457
So far, the functionality you’ve seen has been concerned with presenting data
already stored in the database. Many of the remaining pages involve adding or
updating data.
10.10 Editing a Book
From the MyLibrary page, it is possible to edit an existing Book using one of the
icons on the right side of the page. The HTML template for the MyLibrary page
includes a link that encloses a Rollover component for the button image:
<a jwcid="edit"><img jwcid="editRollover"
src="images/edit.png" width="25" height="25"
alt="Edit" border="0"/></a>
The edit component is a DirectLink. It invokes the editBook() listener method
and encodes the Book’s id as a service parameter:
<component id="edit" type="DirectLink">
<binding name="listener" expression="listeners.editBook"/>
<binding name="parameters" expression="currentBook.id"/>
</component>
Like several other pages, the EditBook page requires an initialization stage
before it can render. This initialization involves reading the Book to be edited.
The EditBook class does not implement the IActivate interface, even though its
needs are similar. The IActivate interface doesn’t leave any room for specifying
the book to edit. The IActivate interface is for pages (like MyLibrary) whose initialization does not require any additional parameters.
Instead, the editBook() listener method, shown in listing 10.28, invokes a
specific method on the EditBook class to begin editing the book identified as a
service parameter.
Listing 10.28 The editBook() method of the MyLibrary class
public void editBook(IRequestCycle cycle)
{
Object[] parameters = cycle.getServiceParameters();
Integer bookId = (Integer) parameters[0];
EditBook page = (EditBook) cycle.getPage("EditBook");
page.beginEdit(cycle, bookId);
}
458
CHAPTER 10
The beginEdit() method, shown in listing 10.29, stores the bookId and reads the
book from the database, using the Operations bean. When the user edits an
existing book here, or adds a new book (using the NewBook page), the book is
stored as a Map of attributes and values, rather than as a Book instance. The Book
object is designed only as the output from the BookQuery bean. It contains flattened data from the Book, Person, and Publisher entities and is not designed
for updates.
Instead, the EditBook page uses a Form component and form control components to edit the values stored in a Map. The Map is then used to update the
Book entity.
Listing 10.29 The beginEdit() method of the EditBook class
public void beginEdit(IRequestCycle cycle, Integer bookId)
{
setBookId(bookId);
int i = 0;
while (true)
{
try
{
IOperations operations = vengine.getOperations();
setAttributes(operations.getBookAttributes(bookId));
break;
}
{
}
{
vengine.rmiFailure(
"Remote exception setting up page for book #" +
bookId + ".",
ex,
i++);
}
}
}
Editing a Book
459
The beginEdit() method uses the Operations bean to get the attributes of the
Book and assigns those attributes to the attributes property of the EditBook
page, ready to be edited by the form control components on the page. Each of
the form control components edits a property within the attributes Map. Tapestry
(or, really, OGNL) treats Maps as if they were JavaBeans, where the properties are
the keys of the Map. When the page renders, each component will read a property
from the attributes Map (such as attributes.title or attributes.author).
What about when the form is submitted? The components will attempt to set
properties of the Map, but since this is a new request, the attributes property will
initially be null. The way to address this is to ensure that a non-null Map is available when the page renders. This is handled by the page’s pageBeginRender()
method, shown in listing 10.30.
Listing 10.30 The pageBeginRender() method of the EditBook class
{
if (getAttributes() == null)
setAttributes(new HashMap());
}
The pageBeginRender() method is invoked before a page renders, but also before
a form within the page rewinds. In either case, this event listener method will
ensure that a non-null Map is stored in the attributes property, ready to be read
or updated by the form control components.
10.10.1 Tracking the Book ID
How do we know which book is being edited? When the page is initially rendered, we know the ID of the book (it is stored in the bookId property), and we
need that piece of information again when the form is submitted. In chapter 4,
we discussed scenarios where clicking the browser’s back button could cause us
some grief (if the user jumps back to previously rendered pages using the
browser’s back button). The solution here is to record the book ID in the form
using a Hidden component. This ensures that the form submission is completely
consistent: The field updates are packaged with the identity of the book to be
updated, and there’s no possibility of the client and the server disagreeing as to
which book is being edited. The following code is included in the EditBook
page’s HTML template as the first thing within the Form component:
<span jwcid="@Hidden" value="ognl:bookId"/>
460
CHAPTER 10
This records the bookId property into the form when rendering, and then
restores the bookId property when the form is submitted. Since the bookId property is not used until the form’s listener is invoked, the Hidden component can
go anywhere on the page, as long as it is enclosed by the Form component.
10.10.2 Generating dynamic JavaScript
The Virtual Library application allows new publishers to be defined on the fly.
When adding a new book or editing an existing book, the user has the option of
either selecting an existing publisher (from a drop-down list) or providing the
name of a new publisher in a text field.
This either-or relationship between the drop-down list and the text field is
enforced in the client web browser using client-side JavaScript (and then doublechecked on the server, in case the client has JavaScript disabled). The JavaScript
consists of client-side event handlers that observe changes to the drop-down list.
When the list has a nonblank value selected, the text field is disabled. When the
user selects the empty option in the drop-down list, the text field is enabled, and
the cursor is moved to the text field. As discussed in chapter 8, JavaScript in Tapestry is complicated by the fact that the framework assigns the names of forms
and form control elements, and this requires a specialized template, the script
specification, to dynamically create JavaScript customized to the names assigned
by the framework.
The EditBook page includes a Script component for creating the JavaScript.
A Script component is another framework component, whose goal is to provide
an easy way to add dynamic JavaScript to a page without our having to resort to
overriding the renderComponent() method (as we did in chapter 8). The script is
passed the two components (informal parameters to the Script component
become input symbols to the script specification), as well as the relative (to the
page specification) location of the script specification. Although the Script component appears in the HTML template, it does not produce any HTML—it works
with the Body component to add JavaScript to the page.
The Script component is configured in the EditBook page’s HTML template
to execute the Publisher.script specification:
<span jwcid="@Script" script="Publisher.script"
select="ognl:components.inputPublisher"
field="ognl:components.inputPublisherName"/>
The Script component reads and executes the Publisher.script file (which is
stored in the WEB-INF folder with the page specification). The two informal
Editing a Book
461
parameters, select and field, are passed to the executable script as input symbols. The script specification is shown in listing 10.31.
Listing 10.31 Publisher.script: script specification used by the EditBook page
3.0//EN"
➥
<script>
<input-symbol key="select"
class="org.apache.tapestry.form.PropertySelection"
required="yes"/>
<input-symbol key="field"
class="org.apache.tapestry.form.AbstractTextField"
required="yes"/>
Receives the
PropertySelection
for the Publisher
Receives the
TextField for the
Publisher name
<let key="formObject">
document.${select.form.name}
</let>
<let key="selectObject">
${formObject}.${select.name}
</let>
<let key="fieldObject">
${formObject}.${field.name}
</let>
<let key="functionName" unique="yes">
onChange_${select.name}
</let>
Defines name of
onchange event
handler function
<body>
function ${functionName}()
{
var select = ${selectObject};
var field = ${fieldObject};
if (select.selectedIndex == 0)
{
field.disabled = false;
field.focus();
field.select();
}
else
{
field.disabled = true;
462
CHAPTER 10
field.blur();
field.value = "";
}
}
</body>
<initialization>
${selectObject}.onchange = ${functionName};
if (${selectObject}.selectedIndex != 0)
${fieldObject}.disabled = true;
</initialization>
</script>
Connects the handler
to the select object
This script specification works by defining new symbols for client-side Document Object Model ( DOM ) references to the <form> , <select> , and <input
type="text"> elements (as symbols formObject, selectObject , and fieldObject). It also constructs a unique name for the event-handling function for the
<select> element.7
The event-handling function appears in the <body> element of the script
specification. The function is invoked when the user has changed the <select>
element to select the first option, empty, and adjusts the text field accordingly.
In the <initialization> element, event-handling functions are connected to
client-side objects. The content of the <initialization> element is executed
once the complete HTML page is loaded into the client web browser. The initialization here sets the onchange event handler for the <select> element to the
event-handling function. It also checks to see if the <select> is displaying a nonempty option and disables the text field if so.
This level of indirection may seem unnecessary in this specific case; you
could just see what names the framework assigns for the fields and write a script
block as static HTML. Doing so would not be a wise idea, however, since a future
version of the framework may change the naming conventions for forms and
form controls.
As a final note, because JavaScript generation in Tapestry is parameterized,
this same script is used in another page: the NewBook page, which also accepts a
publisher name from a drop-down list or a text field.
7
The function name will typically be something like onChange_inputPublisher. This script isn’t reused on the page, so adding unique="yes" to the <let> is overkill, but not a bad habit to get into.
Giving books away
463
The EditBook page has one limitation: It isn’t possible to change the owner
of a book using the page. Instead, a different page (accessible from the menu) is
used: the GiveAwayBooks page, which we discuss next.
10.11 Giving books away
The GiveAwayBooks page (figure 9.18) allows a user to transfer the ownership of
one or more books to another user within the application. This page is built around
a Palette component (to select the books we want to transfer) and a PropertySelection (to select the user we want to transfer books to). What’s interesting is
that the Palette has been configured to display the PropertySelection in the title
area for the right column (normally labeled “Selected”).
This is accomplished using another type of component: a Block. A Block
component is a way of marking a section of a template so that it can be passed to
another component as a parameter. In the GiveAwayBooks template, the form
looks like this:
<select jwcid="selectBooks"/>
<span jwcid="availableTitleBlock@Block">
Books you currently own.
</span>
Contains the
Palette’s available
column title
<span jwcid="selectedTitleBlock@Block">
Books to transfer to <select jwcid="selectTarget"/>
</span>
Contains the
Palette’s selected
column title
<br clear="left"/>
<input type="image" src="images/transfer.png" width="56"
height="20" alt="Transfer"/>
</form>
The selectBooks component is the Palette, and the selectTarget component is
the PropertySelection. In the GiveAwayBook page’s specification, the selectBooks component is configured to replace its default title for the two columns
with the Blocks provided in the template:
<component id="selectBooks" type="contrib:Palette">
<binding name="sort">
@org.apache.tapestry.contrib.palette.SortMode@LABEL
</binding>
464
CHAPTER 10
<binding name="model" expression="booksModel"/>
<binding name="selected" expression="selectedBooks"/>
<binding name="selectedTitleBlock"
expression="components.selectedTitleBlock"/>
<binding name="availableTitleBlock"
expression="components.availableTitleBlock"/>
</component>
The Palette’s availableTitleBlock and selectedTitleBlock parameters are used
to replace the default Blocks (which are inside the Palette’s own template) with
content from the page’s template. As this page demonstrates, this content can
not only include simple text (“Books you currently own”) but can be arbitrarily
complex, with a mix of static HTML and other components.
The GiveAwayBooks page is the last page accessible to ordinary users. The
remaining pages are restricted to users with administrative privilege. Let’s look
at the first of these pages, EditPublishers, next.
10.12 Editing the publishers
The EditPublishers page allows a Virtual Library administrator to rename or
delete any publishers in the Virtual Library database. This is an important and
necessary page, since users may enter publisher names on the fly, and can thus
introduce inconsistencies and misspellings.
This page is an example of a Form that loops over a number of domain
objects, providing form control elements for editing properties of each of the
domain objects. In this case, the domain objects are Publishers, and the property we want to edit is the name of the Publisher. In addition, the form tracks
which Publishers should be deleted. The Operations bean provides a method
for performing all the updates and deletions as a single method invocation.
Chapter 4 discussed the ListEdit component, which is designed for looping
over lists of objects within a form. The EditPublishers page demonstrates how to
properly use the ListEdit component, how to use the ListEditMap utility class,
and how to recover from out-of-date submissions (which can happen when the
user hits the browser’s back button and resubmits a form).
10.12.1 Constructing the EditPublishers template
The template for the EditPublishers page contains a Form component, a ListEdit
component (the e component), a TextField for editing the publisher’s name property, and a Checkbox for marking publishers as deleted. The EditPublishers
template is shown in listing 10.32.
Editing the publishers
465
Listing 10.32 EditPublishers.html: HTML template for the EditPublishers page
<body jwcid="@Border">
<span jwcid="@ShowMessage"/>
<span jwcid="@ShowError"/>
<form jwcid="@Form" listener="ognl:listeners.processForm">
<p>This page allows you to rename and delete publishers.
</p>
<table class="data">
<tr>
<th>Publisher</th>
<th>Delete</th>
Iterates over the
</tr>
available Publishers
<tr jwcid="listEdit">
<td class="control">
<input jwcid="inputName@TextField"
Edits current
Publisher’s name
value="ognl:publisher.name"
property
size="40" maxlength="40"/>
</td>
<td class="checkbox">
<input type="checkbox"
Edits listEditMap’s
jwcid="delete@Checkbox"
deleted property
selected="ognl:listEditMap.deleted"/>
</td>
</tr>
<tr>
<td class="control">
<input type="image" src="images/update.png"
width="52" height="20" alt="Update"/>
</td>
</tr>
</table>
</form>
</body>
</html>
The checkbox does not edit a property of the Publisher object ( Publisher
doesn’t have any kind of deleted property). Instead, it edits the property listEditMap.deleted. As you’ll see shortly, the ListEditMap object is essential in
terms of handling the form submission, including its ability to track the keys of
objects that should be deleted.
10.12.2 Declaring properties for the EditPublishers page
The page specification for the EditPublishers page (listing 10.33) declares the
standard message and error properties. It also declares a publisher property
466
CHAPTER 10
(which stores the current Publisher instance inside the loop) and a listEditMap
property, used in concert with the ListEdit component to convert between publisher IDs and Publisher objects and to track which publishers should be deleted.
The ListEdit component was previously discussed in chapter 4, and the EditPublishers page is a perfect example of how to make use of it.
Listing 10.33 EditPublishers.page: specification for the EditPublishers page
<page-specification
class="org.apache.tapestry.vlib.pages.admin.EditPublishers">
<property-specification name="listEditMap"
type="org.apache.tapestry.form.ListEditMap"/>
<property-specification name="message" type="java.lang.String"/>
<property-specification name="error" type="java.lang.String"/>
<property-specification name="publisher"
type="org.apache.tapestry.vlib.ejb.Publisher"/>
<component id="listEdit" type="ListEdit">
<binding name="source" expression="listEditMap.keys"/>
<binding name="value" expression="listEditMap.key"/>
<static-binding name="element" value="tr"/>
<binding name="listener"
expression="listeners.synchronizePublisher"/>
</component>
The ListEdit component sets the key property of the ListEditMap instance on
each iteration. It is up to the synchronizePublisher() listener method (which is
invoked by the ListEdit component on each pass through its loop) to convert this
to a Publisher. Listing 10.34 shows the synchronizePublisher() method.
Listing 10.34 The synchronizePublisher() method of the EditPublishers class
public void synchronizePublisher(IRequestCycle cycle)
{
Publisher publisher = (Publisher) map.getValue();
467
if (publisher == null)
{
setError(getMessage("out-of-date"));
throw new PageRedirectException(this);
}
setPublisher(publisher);
}
When the form is submitted, a new ListEditMap instance will be created from the
data stored in the database at that time. If other users have been editing publishers, the data in the new ListEditMap may not precisely match the data in the
ListEditMap used when the page was rendered. The ListEdit component will pull
the publisher IDs from the form submission, but it will still plug each successive
publisher ID into the Map’s key property, and it will still invoke the ListEdit’s listener method to synchronize the publisher property.
If two users are both deleting Publishers at the same time, then the second
user’s form submission will include publisher IDs that were deleted by the first
user. In that case, the ListEditMap instance will return null from its getValue()
method for that publisher ID. That’s why the synchronizePublisher() method
includes a null check—null indicates that just such a race condition has occurred
and that the user is out of date. Throwing a PageRedirectException aborts the
form rewind and renders the EditPublishers page again.
10.12.3 Creating the ListEditMap
A ListEditMap instance is created whenever the page renders. As usual, this is
accomplished using the PageRenderListener interface. Listing 10.35 contains the
short pageBeginRender() method, as well as the longer readPublishers() method.
Listing 10.35 The pageBeginRender() method of the EditPublishers class
{
readPublishers();
}
private void readPublishers()
{
Publisher[] publishers = null;
468
CHAPTER 10
int i = 0;
while (true)
{
try
{
publishers = operations.getPublishers();
break;
}
{
vengine.rmiFailure(getMessage("read-failure"), ex, i++);
}
}
ListEditMap map = new ListEditMap();
int count = Tapestry.size(publishers);
for (i = 0; i < count; i++)
map.add(publishers[i].getId(), publishers[i]);
setListEditMap(map);
}
The readPublishers() method uses the Operations bean to get the list of publishers, sorted in alphabetical order. It uses this list to construct the ListEditMap
instance, which is stored in the listEditMap property of the page.
A ListEditMap instance remembers the order in which keys are added to it.
The keys property, used by the ListEdit component as the source of values
when rendering, will return the keys in this order. The publishers are displayed in ascending alphabetical order because the Operations bean’s getPublishers() method returns them in that order, and that is the order in which
they are added to the ListEditMap.
The ListEditMap’s getValue() method doesn’t require a parameter to indicate which value to get. The ListEdit component will have already updated the
key property of the ListEditMap for this pass through its loop—that is, the key
used by the getValue() method. The listener method obtains this value and casts
it to type Publisher.
469
10.12.4 Updating the publishers
When the form is submitted, the ListEditMap instance is re-created from current
data stored in the database (the pageBeginRender() method is invoked before the
form rewinds as well as before the page renders). As the form is rewound, the
Publisher objects are updated, and any deleted publishers are tracked.
Inside the form’s processForm() listener method (provided in listing 10.36),
the updates are organized. The ListEditMap’s getValues() method returns all
the values (in this case, Publisher objects) that are not deleted. The getDeletedKeys() method returns the IDs of all the deleted publishers. Care must be taken
because the ListEditMap returns null from getDeletedKeys() when no publishers
have been marked for deletion.
Listing 10.36 The processForm() method of the EditPublishers class
public void processForm(IRequestCycle cycle)
{
if (isInError())
return;
List updateList = map.getValues();
List deletedIds = map.getDeletedKeys();
Publisher[] updated = (Publisher[])
updateList.toArray(new Publisher[updateList.size()]);
Integer[] deleted =
deletedIds == null
? null
: (Integer[])
deletedIds.toArray(new Integer[deletedIds.size()]);
int i = 0;
while (true)
{
try
{
operations.updatePublishers(updated, deleted);
break;
}
470
CHAPTER 10
{
}
catch (RemoveException ex)
{
}
{
vengine.rmiFailure(getMessage("update-failure"), ex, i++);
}
}
vengine.clearCache();
}
After the values and deleted IDs are extracted and converted to arrays, the Operations EJB is again used to perform the update as a single operation. After the
normal retry logic, the final step is to have the engine clear any cached information it might have (the engine caches a list of publishers).
The same techniques used by the EditPublishers page are expanded for use
on the EditUsers page.
10.13 Editing the list of users
Like the EditPublishers page, the EditUsers page is accessible only to administrators. It allows the current user to edit select aspects of the other users: granting or revoking administrative privilege, locking or unlocking users, changing
users’ passwords, or deleting users outright.
As with deleting publishers (on the EditPublishers page), tracking which
users will have their password changed is a function of the ListEditMap rather
than an attribute of the Person object. This requires a subclass of ListEditMap.
The EditUsers page’s HTML template contains the same kind of loop, built
around a ListEdit component, as the EditPublishers page. The checkbox for the
reset password column is bound to the ListEditMap’s resetPassword property:
<tr jwcid="listEdit">
value="ognl:user.naturalName">Joe User</span>
<br>
<span jwcid="@Insert"
value="ognl:user.email">[email protected]</span>
</td>
471
jwcid="inputAdmin@Checkbox"
selected="ognl:user.admin"/>
</td>
jwcid="inputLockedOut@Checkbox"
selected="ognl:user.lockedOut"/>
</td>
jwcid="inputResetPassword@Checkbox"
selected="ognl:listEditMap.resetPassword"/>
</td>
jwcid="inputDelete@Checkbox"
selected="ognl:listEditMap.deleted"/>
</td>
</tr>
10.13.1 Creating the ListEditMap subclass
The purpose of the UserListEditMap is to extend ListEditMap to also track a
set of keys identifying which Person values should have their passwords reset.
This involves creating a Set to track those keys and making use of methods
provided by the ListEditMap super class to use that Set. ListEditMap is designed
specifically to support this kind of extension, so the final class (listing 10.37) is
quite succinct.
Listing 10.37 UserListEditMap.java: ListEditMap subclass used on the EditUsers page
package org.apache.tapestry.vlib.pages.admin;
import java.util.Set;
import org.apache.tapestry.form.ListEditMap;
public class UserListEditMap extends ListEditMap
{
private Set _resetPasswordKeys;
public List getResetPasswordKeys()
{
return convertSetToList(_resetPasswordKeys);
}
472
CHAPTER 10
public boolean getResetPassword()
{
return checkSet(_resetPasswordKeys);
}
public void setResetPassword(boolean resetPassword)
{
_resetPasswordKeys =
updateSet(_resetPasswordKeys, resetPassword);
}
}
The ListEditMap class provides three protected methods to be leveraged by subclasses: checkSet() , updateSet(), and convertSetToList() . The checkSet()
method checks a Set to see if it contains the current key. It automatically returns
false if the Set is null.
The updateSet() method updates a Set, adding or removing the current
key based on the boolean parameter. It may seem odd that the method takes a
Set as a parameter and returns the same Set as a result. This is because the Set
passed in may be null. In that case, updateSet() will create a new instance of
Set and return it. This slight twist of the code allows the Set to be created only
if it is needed.
Finally, the convertSetToList() method converts a Set (possibly null) to a
List. It can return null if the set is null or empty. The order of elements in the
returned List is not defined. In this case, List is returned because access to
the values inside a List is more efficient than with a Collection. With a List,
it is possible to get each element using the get() method (passing in an
index). With a Collection, it is possible to access all elements only with the use
of an Iterator.
10.13.2 Handling the form submission
When the form is submitted, the UserListEditMap instance is used to get the
updated users, as well as the list of deleted user IDs and the list of user IDs
whose password will be reset. Listing 10.38 contains the updateUsers() listener
method and a utility method, toArray(), used to convert a Collection to an
array of objects.
473
Listing 10.38 The updateUsers() and toArray() methods of the EditUsers class
public void updateUsers(IRequestCycle cycle)
{
if (isInError())
Skips update if
validation errors
return;
(VirtualLibraryEngine) cycle.getEngine();
UserListEditMap map = getListEditMap();
List updatedUsers = map.getValues();
Person[]updates = (Person[])
updatedUsers.toArray(
new Person[updatedUsers.size()]);
Gets Person
objects to
update
Gets IDs of Persons
Integer[] resetPasswordUserIds =
to reset passwords
toArray(map.getResetPasswordKeys());
Integer[] deletedUserIds =
Gets IDs of
toArray(map.getDeletedKeys());
Persons to delete
String password = getPassword();
setPassword(null);
Gets and clears
password field
if (Tapestry.isBlank(password) &&
Tapestry.size(resetPasswordUserIds) != 0)
{
setErrorField("inputPassword", getMessage("need-password"));
return;
}
Integer adminId = visit.getUserId();
int i = 0;
while (true)
{
try
{
operations.updatePersons(
updates,
resetPasswordUserIds,
password,
deletedUserIds,
adminId);
break;
Performs all
updates and
deletes
}
474
CHAPTER 10
{
vengine.rmiFailure(getString("update-failure"), ex, i++);
}
catch (RemoveException ex)
{
}
{
}
}
setMessage(getMessage("users-updated"));
}
private Integer[] toArray(Collection c)
{
int count = Tapestry.size(c);
if (count == 0)
return null;
return (Integer[]) c.toArray(new Integer[count]);
}
With that, we’ve covered all the key pages and the components in the application—but that’s not quite the entire application. We still need to define the web
deployment descriptor, and there a few interesting deployment issues there.
10.14 Creating the web deployment descriptor
A common desire for web applications is that the URL be short and succinct;
this is as much for marketing reasons as for technical reasons, and is perfectly
valid. Users should just need to know the domain name, such as http://www.
amazon.com or http://www.slashdot.org, in order to launch the application.
Tapestry applications, however, require a reference to the application servlet, so
the URL for launching a Tapestry application has an extra term—for example:
http://www.myserver.com/app. Let’s see how to avoid that unwanted part, /app,
at the end.
When using a JSP application, it is possible to specify a JSP as the “welcome
page.” The servlet container is responsible for recognizing a “bare” URL (a URL
that references a folder, not a particular JSP) and will send a redirect response to
the client, which results in a new request for the welcome page JSP. When Servlet
Creating the web deployment descriptor
475
API 2.4 becomes widely available, it will be possible to perform this same kind of
redirection to a servlet.
In the meantime, Tapestry makes use of a feature of the 2.3 Servlet API:8
Servlet filters. A filter is “plugged into” the request-processing pipeline within
the servlet container. The RedirectFilter class provided with the framework
intercepts the initial request for the bare URL and sends a client redirect response,
forwarding the client to the application servlet.
To make use of this feature, we must declare the filter in the web.xml deployment descriptor, much as we define the Tapestry application servlet. This is
shown in listing 10.39. Because filters are part of the Servlet 2.3 API, we must use
the <!DOCTYPE> for the 2.3 API (in previous examples, we used the 2.2 DOCTYPE).
Listing 10.39 web.xml: deployment descriptor for the Virtual Library application
<!DOCTYPE web-app PUBLIC
"-//Sun Microsystems, Inc.//DTD Web Application 2.3//EN"
"http://java.sun.com/dtd/web-app_2_3.dtd">
<web-app>
<display-name>Tapestry Virtual Library Demo</display-name>
<filter>
<filter-name>redirect</filter-name>
<filter-class>
org.apache.tapestry.RedirectFilter
</filter-class>
</filter>
Defines the
redirect filter
<filter-mapping>
<filter-name>redirect</filter-name>
<url-pattern>/</url-pattern>
</filter-mapping>
Maps the
redirect filter
to bare URL
<servlet>
<servlet-name>vlib</servlet-name>
<servlet-class>
org.apache.tapestry.ApplicationServlet
</servlet-class>
</servlet>
<servlet-mapping>
<servlet-name>vlib</servlet-name>
8
Other than this, Tapestry uses only features of the Servlet 2.2 API, for maximum portability to the
many application servers in use even today.
476
CHAPTER 10
</servlet-mapping>
<session-config>
<session-timeout>15</session-timeout>
</session-config>
</web-app>
That shortens the launch URL for the application from http://www.myserver.com/
vlib/app down to http://www.myserver.com/vlib. Ideally, the launch URL should
be just http://www.myserver.com. Going that extra step takes two different forms,
depending on how the application is deployed.
10.14.1 Deploying web applications as root
When you’re deploying an individual web application, the name of the WAR file
becomes the name of the prefix. So, a web application named demo.war would
be deployed as http://www.myserver.com/demo. Most servlet containers, including Jakarta Tomcat, follow a convention that a web application named ROOT.war
is deployed without a prefix, as http://www.myserver.com/. Combining such a
deployment with the filter described earlier allows us to reach that ideal, minimallaunch URL: http://www.myserver.com.
10.14.2 Deploying an enterprise application as root
The Virtual Library does not deploy as a simple web application; it is deployed
as an enterprise application. An enterprise application is another level above a
web application and bundles together one or more web applications with any
number of other libraries, including EJB modules.
In the case of the Virtual Library, the vlib.war web application is combined
with the vlibbeans.jar EJB module to form the vlib.ear enterprise application.
vlibbeans.jar and vlib.war are built separately and then combined to form the
final application. These two modules are included inside the enterprise application archive (which is another version of a JAR file, with a different extension, .ear).
Just as a web application has a web.xml deployment descriptor and an EJB
module has an ejb-jar.xml deployment descriptor, an enterprise application has
an application.xml deployment descriptor. The descriptor for the Virtual Library
enterprise application is shown in listing 10.40.
Wrapping it all up
477
Listing 10.40 application.xml: enterprise application deployment descriptor for
the Virtual Library application
"-//Sun Microsystems, Inc.//DTD J2EE Application 1.2//EN"
"http://java.sun.com/j2ee/dtds/application_1_2.dtd">
<application>
<display-name>Virtual Library Application</display-name>
<module>
<ejb>vlibbeans.jar</ejb>
</module>
<module>
<web>
<web-uri>vlib.war</web-uri>
<context-root>/</context-root>
</web>
</module>
</application>
The application deployment descriptor identifies how to deploy the two modules packaged within the enterprise application archive, identifying one as an
EJB module and the other as a web application. The critical part in this discussion is the <context-root> element. This element gives us absolute control over
the URL prefix used to identify the web application; the name of the web application archive is no longer relevant.
For our purposes, we deploy the web application as / rather than /vlib. The
end result is that the application will launch with the URL http://www.myserver.com, which is what we want.
10.15 Wrapping it all up
Every application contains a unique set of challenges driven by the need to know
who the user is, what the user is doing, and what data is stored on the server (as
persistent page properties) or in the client (as service parameters or hidden
form fields). Tapestry provides a clean, object-oriented approach to meeting all
these challenges.
You’ve seen that it is possible, in a rather small amount of code, to create a
complex, robust, usable application. Tapestry provides the necessary hooks, in
the form of page lifecycle events, to fill in any gaps in functionality that the
framework can’t provide. All of these techniques can be expanded and adapted
to new applications with new challenges.
478
CHAPTER 10
In the space of these few chapters, you’ve seen how Tapestry simplifies and
improves web application development in everything from a simple word game
to an entire J2EE application. Tapestry accomplishes quite a bit from just a few
basic building blocks:
■
Tapestry’s component object model allows the framework to manage the
lifecycle of pages and components and facilitates the ability to address a
particular page or component within a URL.
■
Expressing the relationships and interactions within the framework in
terms of JavaBean properties eliminates the impedance between different
parts of the application—especially between the components provided by
the framework and the domain objects specific to your application.
■
Dividing an application into self-contained pages and achieving reuse
through the creation of components leads to more functionality and fewer
bugs, and requires less code.
■
Embracing a separation of concerns by dividing pages (and components)
into Java classes, HTML templates, and specifications mirrors the ModelView-Controller pattern and enables developers to work together easily.
This book has given you the tools to understand how to create your own custom
Tapestry applications. You are well on your way to thinking in terms of components: reusing existing components or creating your own (with an eye toward
reusing them). This book, and these examples, are just the start of the Tapestry
world, which begins at the Tapestry home page: http://jakarta.apache.org/tapestry/. Tapestry is built around a thriving, world-spanning community of users and
volunteers, which translates to a wealth of support and examples.
Using Tapestry, you are freed from the most troublesome and taxing concerns
of developing web applications. Now you can concentrate on the real challenges:
unrealistic project schedules, difficult clients, the changing technology landscape, and all the other realities of twenty-first century software development.
“OK, folks, we don’t get much of a break, the client is really breathing down our necks for
phase two, and we’ve got a lot of work to do.” As usual, you tune out for a bit and start visualizing the changes; the new registration wizard is twisted, may be time to refine those workflow
components you’ve been prototyping … you’ll need a component to raise that new pop-up
window … and you’ll have to whip something together for that Flash animation. You frown.
The new functionality is going to break the fragile links into the client’s back-end systems.
Again. More quality time with their IT department. Why couldn’t it all be as simple and
straightforward as the web interface?
Getting involved
with Tapestry
479
480
APPENDIX A
One of the great aspects of any open-source project is just how easy it is to get
started and, potentially, guide the evolution of the project. The starting point for
participation is the Tapestry home page at http://jakarta.apache.org/tapestry/ and
the two mailing lists, [email protected] (where you’ll find discussions on how to use Tapestry) and [email protected] (which focuses
on ways to improve the framework). Instructions for subscribing to the mailing
lists are available on the Tapestry home page.
Tapestry has benefited immensely from the ideas brought up for discussion in
the mailing lists. For example, line-precise error reporting evolved from a strong
suggestion (really more of a complaint) by a new user who had difficultly working
backward from runtime errors to mistakes in the XML specifications. That simple
request led to significant improvement to the framework.
Here’s another example: Through Tapestry release 2.3, there was no concept
of an implicit component; that came out of a discussion involving Marc Fleury
(the volatile lead of the JBoss application server project). Again, a major enhancement to the framework came out of direct discussions with the framework’s end
users. In many more cases, extensions to individual components or to APIs were
a result of end-user feedback.
If you want to go beyond making suggestions to actually supplying code,
there is a path for this as well. Tapestry is an Apache Jakarta project and is organized along the Apache meritocracy guidelines—which means all decisions are
made out in the open, on the mailing lists, and are based on open votes. There
are no barriers to entry, beyond an ability to communicate, a willingness to work
with the Tapestry team, and, of course, the ability to produce high-quality code.
Several current members of the Tapestry team came up through the ranks by
mentoring new users on the mailing lists and eventually submitting code patches,
before being voted in as Tapestry committers.
A.1 Tacos—the Tapestry component repository
As you saw in chapters 6 and 8, Tapestry has sophisticated support for component
libraries. Creating, packaging, and using a Tapestry component library is quite
simple, and Tacos, the Tapestry Component Archive project, is a source for these
libraries. Tacos is a SourceForge project located at http://sf.net/projects/tacos/. At
the time of this writing, Tacos is in a formative stage, but plans are in progress to
make it the second epicenter for Tapestry development (after the Tapestry home
page at Jakarta) and a source for Tapestry components, sample applications, and
application skeletons.
What’s coming next in Tapestry?
481
A.2 What’s coming next in Tapestry?
Tapestry is an ongoing project at Jakarta. At the time of this writing, Tapestry 3.0
is deep into its beta period, but advanced users have helped the team identify a
number of areas deserving of attention in the next full release.
By the time this book is published, work will be well under way on release 3.1
of Tapestry. The community is very active, and there is a constant drive to
improve the framework—a continuous search for ways to extend the power of
Tapestry and, at the same time, reduce even further the time and effort involved
in creating Tapestry applications.
A.2.1 Improving application testability
Development and testing go hand in hand. If you want to complete a project on
time, your best bet is to take the time to develop tests for your code in parallel
with the code itself. Some developers, those who embrace the Extreme Programming philosophies, go further: They write the tests first and then create code to
make the tests pass.
This seems like a paradox: If you are under the gun, struggling to meet an
aggressive development schedule, writing tests seems like a luxury you can’t
afford. That’s a terrible misconception. The minutes you may spend creating a
test suite are a small price to pay compared with the hours you can waste tracking down bugs deployed into the bowels of an application server. Think about
it—inside a unit-test suite, you have a simple, predictable environment where it
is easy and natural to use the debugger. Running your test suite should be as
simple as hitting the Test button in your IDE.
Finding a bug inside your application server is much more complicated; you
have to deal with debugging a remote process, with building and deploying the
application, and with filtering through all of the application server overhead. In
short, it’s a long and complicated cycle, adding minutes or more to each iteration of your test cycle.
As we’ve often said, when you write tests, you find bugs. Those bugs will eventually come to light in integration testing of the application, long after you’ve forgotten the minute details of how your code works—which will further multiply
the time and effort needed to find and fix the bugs. The only solution is to create
tests early.
By dividing up your code carefully, you can test portions of your application
outside the Tapestry framework; for example, back in chapter 2, the Game and
WordSource classes could be tested easily using their own unit-test suites. These
482
APPENDIX A
classes were carefully coded to have no dependencies on Tapestry or on any
other user interface.
Unfortunately, the majority of a Tapestry application requires the active
involvement of the Tapestry framework to be tested. Not only do most of the
component classes fail to work in isolation, but many of them are abstract classes
that can’t even be instantiated, much less tested. Tapestry applications can only
be tested with the Tapestry framework.
The testing story for Tapestry applications is less than ideal:
■
Obtain a testing tool, such as Mercury Interactive’s LoadRunner, Minq
Software’s PureLoad, or OpenSTA.1
■
Set up a test environment for the tool: an application server and any backend database or other external dependencies.
■
Build the application and deploy it to the application server.
■
Use the tool to execute the tests.
■
Analyze the results.
That’s not a bad scenario for performing load and performance testing, but it is
not appropriate for day-to-day (or even minute-to-minute) functional testing. As
a developer, you want to hit that Test button to run your application through its
paces to prove you haven’t broken anything, and you don’t want to wait long. You
want to use the same build-and-test strategy you’d use for a noninteractive or
nonweb application.
A number of extensions to the JUnit framework are available for merging unit
testing with web application testing. You’ll find these listed at http://www.
junit.org/news/extension/web/index.htm. All of these extensions require a running web server or application server to execute against; the excellent opensource servlet container Jetty (http://jetty.mortbay.com) is easily embeddable for
testing purposes.
Internally, Tapestry includes an extensive test suite in excess of 400 tests. A
significant number of these tests are based on executing small applications
inside a simulated servlet container. These tests execute several hundred simulated HTTP requests in the space of a few seconds.
A focus for Tapestry release 3.1 will be to improve, generalize, and document
the internal test suite, converting it into a part of the Tapestry framework proper.
1
Information about web site testing software is available online at www.extremetech.com/article2/
0,3973,1154892,00.asp, http://www.aptest.com/resources.html#web-func, and elsewhere.
What’s coming next in Tapestry?
483
This will provide a starting point for you to build a test suite for your own application, allowing you to perform constant, pervasive testing without the complexity and overhead of running a servlet container or application server.
A.2.2 Offline content generation
There’s no question that static content can be served up by a web server at tremendous speed; static content can be held in memory, cached in multiple locations, and more easily distributed than dynamic content. As soon as the content
is dynamic, there is at least an order of magnitude drop-off in speed, and this
only gets worse when the application is database driven.
Large content sites, such as Yahoo! and CNN, take the approach of generating
static content at frequent intervals (every hour, or even every few minutes). This
splits the difference between a fully dynamic, fully personalized web site, and a
fast, static web site. Various contributors to the Tapestry development mailing
lists have described approaches they have taken to bridge between the static and
dynamic approaches. One approach is to keep a Tapestry application running
and periodically use scripts (written in Perl or Python) to “scrape” the dynamic
content and store it to a static file, accessible via a web server.
For Tapestry to be useful in the very highest top-end applications, the kind
visited by thousands of concurrent users, offline content generation is at least as
important as application server clustering. Offline content generation will be
another key goal in Tapestry 3.1.
A.2.3 Other simplifications and improvements
Part of the fun of developing Tapestry is that there’s always something new to be
added, refined, or sometimes rethought. Discussions with users on the Tapestry
mailing lists (see the next section) can be quite invigorating.
There are at least two additional areas in Tapestry that will receive some
attention in release 3.1.
Simplifying parameters
The use of the <parameter> direction attribute to describe how Tapestry component parameters are used (covered in chapter 6) is less than ideal. It requires that
you, the developer, understand exactly how data moves between a component
and its container (whether the container is a page or another component). The
idea was to optimize the number of accesses to the binding object so that properties would not be read or updated unnecessarily. These accesses, both reads and
writes, rely on Java reflection, which is much slower than ordinary Java method
invocation, so optimizations here can be very worthwhile.
484
APPENDIX A
With the bytecode enhancement technology introduced in release 3.0, it is
quite possible for a much improved version of the auto value for the direction
attribute to be implemented that can accomplish the same, or better, optimizations with even greater efficiency.
Simplifying persistent properties
As you saw in chapter 10, sometimes elaborate measures must be taken to work
around the fact that persistent page properties may not be modified once the
page begins rendering. Release 3.1 will explore whether these strict rules surrounding persistent page properties may be relaxed.
Building the
examples with Ant
485
486
APPENDIX B
Ant is an open-source build tool for Java that is used in a vast number of opensource and proprietary software projects. Ant is used to automate the steps for
building, testing, and deploying Java software. The examples used throughout this
book are available in source form from the Manning web site for this book,
www.manning.com/lewisship/. This appendix describes how to use Ant to convert
that source into applications that run within the Tomcat servlet container.
A complete, indispensable, and definitive guide to Ant is Java Development
with Ant, by Erik Hatcher and Steve Loughran (Manning, 2002).
B.1 Downloading the software
Building the example applications involves four software packages:
■
The Ant tool, which will compile the classes and assemble and deploy the
web applications.
■
The Tomcat servlet container. Tomcat is the reference implementation of
the Java Servlet API.
■
The Tapestry framework.
■
The Tapestry in Action examples
Table B.1 lists the home page for each package (you’ll see a download link on
the page) and specifies which version you should use.1
Table B.1
Packages ued to build the examples
Package
Home page
Version
Ant
http://ant.apache.org
1.5.1
Tomcat
http://jakarta.apache.org/tomcat/
4.1.24
Tapestry
http://jakarta.apache.org/tapestry/
3.01
Tapestry in Action examples
www.manning.com/lewisship/
N/A
B.1.1 Installing Ant
Unpack the Ant distribution (it will be a zip file for Windows, or a .tar.gz file for
GNU/Linux or UNIX) to a directory. Ant includes installation directions in the
docs/manual/index.html file. The installation process involves updating the
1
At the time of this writing, Tapestry 3.0 was in a late beta stage. By the time you read this, Tapestry
will have reached its final 3.0 release.
Downloading the software
487
system PATH environment variable, adding Ant’s bin directory to the list of
search directories. The bin directory contains scripts that run Ant. In addition,
you will set up an ANT_HOME environment variable to point to the installation location of Ant.
B.1.2 Installing Tomcat
On Windows, Tomcat is easiest to install by downloading as a self-extracting
archive (.exe) file, which has an integrated installer. By default, the server is
installed in the directory C:/Program Files/Apache Group/Tomcat 4.1. For GNU/
Linux and UNIX systems, a .tar.gz file may be extracted instead.
B.1.3 Installing Tapestry
Tapestry is distributed as a zip file for Windows and as a .tar.gz file for GNU/
Linux and UNIX. The binary distribution includes the compiled libraries as well
as a number of dependencies needed by Tapestry (these are other open source
libraries that Tapestry makes use of, such as OGNL).
You should extract the Tapestry distribution to a working directory.
Tapestry includes a README file with additional instructions for configuring
its turnkey demos for use with Tomcat or with JBoss.
B.1.4 Installing the examples
Installation of the examples involves two steps. First the archive file, TapestryInActionSource.zip, must be extracted to a working directory.
Second, you must create a file, common/build.properties (listing B.1), within
the working directory. This file is referenced by the Ant build script to identify
where Tapestry and Tomcat have been installed. Listing B.1 is an example of this
file for a Windows installation. The distribution includes a sample file, build.
properties.template, which you can edit to match your choice of directories and
then rename to build.properties.
Listing B.1 build.properties: configuration file for building the examples
tapestry.dist.dir = C:/Tapestry-3.0
tomcat.dir = C:/Program Files/Apache Group/Tomcat 4.1
488
APPENDIX B
B.2 Building the examples
Now that you’ve extracted the examples and created the configuration file, Ant
can compile and deploy the four web applications directly into Tomcat. Simply
execute the command ant from the examples distribution directory.
The top-level Ant build file will, by default, build the library, and then build
and deploy each of the four web applications.
In addition, a small change to the Tomcat configuration is installed. Its goal
is to resolve a naming conflict between the Tapestry examples web application
and an example web application distributed as part of Tomcat.
B.3 Running the examples
You should start Tomcat and allow it to fully initialize. On Windows, the installer creates a Start menu group; on other systems, follow the directions in the distribution.
Once Tomcat has initialized and loaded the applications, you may start a web
browser and open the URL http://localhost:8080/examples/app. This will display
the Home page of the main examples application, which has links to all the
pages of the examples application and to the other three applications (both versions of Hangman and the banner ads demo).
B.4 Understanding the examples build system
The examples distribution consists of a number of subprojects, each in its own
subfolder, and each with its own Ant build file. These subprojects are shown in
table B.2. A master Ant build file, in the distribution directory, links together all
the subprojects.
Table B.2
Subprojects in the examples distribution
Name
Type
Description
examples-library
Library
Library containing components described in chapters 7 and 8
examples
Web application
Application containing pages and components from chapters 3–7
hangman1
Web application
Application described in chapter 2
hangman2
Web application
bannerads
Web application
Understanding the examples build system
489
Figure B.1
The master build file
calls into the build
files in each of the five
subprojects (only
hangman1 is shown).
The build environment used by the examples is worth taking a look at as a starting point for building your own build environment. The basic layout is shown in
figure B.1 (which shows only the hangman1 subproject).
Each file and directory has a specific purpose during the build, as table B.3 shows.
Table B.3
Files and directories for the hangman1 subproject
File
Purpose
build.xml
Master Ant build file for the entire project
common/build.properties
Identifies the location of the Tapestry and Tomcat distributions
hangman1/build.xml
Rules for building just the hangman1 web application
hangman1/src/java
Java code to be compiled
hangman1/target/classes
Compiled Java source code (created as needed)
490
APPENDIX B
Table B.3
Files and directories for the hangman1 subproject (continued)
File
Purpose
hangman1/context
HTML templates for pages, plus static assets
hangman1/context/WEB-INF
web.xml deployment descriptor, page and component specifications,
and so on
hangman1/target/hangman1.war
Completed WAR, including compiled Java code, contents of context,
plus all necessary Tapestry libraries and dependencies
In each subproject is a build.xml file used by Ant and an src directory. Ant will
create the target directory (which contains all the derived files, such as compiled
Java classes) and ultimately create the deployable web application archive within
the target directory.
Listing B.2 shows the build file for the hangman1 subproject.
Listing B.2 build.xml: Ant build file for the hangman1 subproject
<project name="Tapestry In Action -- Hangman, version 1"
default="war">
<property name="project.name" value="hangman1"/>
Reads
configuration file
<property file="../common/build.properties"/>
<property name="target.dir" value="target"/>
<property name="src.dir" value="src/java"/>
<property name="context.dir" value="src/context"/>
<property name="conf.dir" value="src/conf"/>
<property name="classes.dir" value="${target.dir}/classes"/>
<property name="war.file"
value="${target.dir}/${project.name}.war"/>
<property name="servlet.jar"
value="${tomcat.dir}/common/lib/servlet.jar"/>
<path id="project.class.path">
<fileset dir="${tapestry.dist.dir}/lib">
<include name="*.jar"/>
<include name="ext/*.jar"/>
</fileset>
<pathelement location="${servlet.jar}"/>
</path>
Constructs
compile
classpath
Defines
symbols
for files
and
folders
Defines path
to Servlet API
library
Understanding the examples build system
491
<target name="clean" description="Deletes derived files.">
<delete dir="${target.dir}" quiet="true"/>
</target>
<target name="compile" description="Compile Java classes.">
<mkdir dir="${classes.dir}"/>
<javac srcdir="${src.dir}" destdir="${classes.dir}" debug="on"
classpathref="project.class.path"/>
</target>
<target name="war" depends="compile"
description="Compile all classes and build the WAR.">
<war warfile="${war.file}"
webxml="${context.dir}/WEB-INF/web.xml">
<fileset dir="${context.dir}">
<exclude name="WEB-INF/web.xml"/>
</fileset>
<classes dir="${classes.dir}"/>
<classes dir="${src.dir}">
<exclude name="**/*.java"/>
<exclude name="**/package.html"/>
</classes>
Packages
application
as WAR file
<lib dir="${tapestry.dist.dir}/lib">
</lib>
<lib dir="${tapestry.dist.dir}/lib/ext">
</lib>
</war>
</target>
<target name="deploy" depends="war"
description="Deploy the WAR into Tomcat.">
<property name="deploy.dir" value="${tomcat.dir}/webapps"/>
<copy file="${war.file}" todir="${deploy.dir}"/>
</target>
</project>
The compile target is responsible for compiling the Java source code, stored
under src/java. The class files are stored under target/classes. The WAR file is created by combining these compiled classes with the image of the web application
context, plus any additional artifacts inside the source directory (this includes
page and component specifications, HTML templates, script specifications, or
492
APPENDIX B
other assets). In addition, the Tapestry libraries and dependencies are copied
into the WAR so that they will be available at runtime.
The last step is to copy the WAR to the Tomcat webapps directory so that
Tomcat can serve the web application.
Tapestry component
reference
493
494
APPENDIX C
Tapestry component reference
This appendix is a quick reference to the basic components provided with the Tapestry framework. A more complete reference, with examples, is provided online and
as part of the Tapestry distribution. We cover only the framework components here.
The framework includes two other libraries: the Contrib library, which includes
recent contributions to the framework (including the Palette component discussed
in chapter 6), and a set of WML (Wireless Markup Language) components.
These reference materials briefly identify each component and its primary
formal parameters (more obscure parameters are omitted for brevity; the online
reference has complete details). In addition, each component can either render
or discard its body, and can either accept or reject informal parameters.
ActionLink
The ActionLink component creates an <a> hyperlink. When the link is triggered,
a listener method is invoked. ActionLink differs from DirectLink in that it uses
the action service, not the direct service, to rewind the entire state of the page
back to its incarnation when the link was initially rendered. (This is the same
technique used by forms to rewind the state of the form submission, as discussed
in chapter 3.)
Table C.1
Parameters for the ActionLink component
Parameter
name
Type
Direction
Required
Default
value
false
disabled
boolean
in
No
listener
IActionListener
in
Yes
Description
If true, the ActionLink
doesn’t render a tag (but still
renders its body).
The listener to be invoked
when the link is triggered.
ActionLink renders its body and allows informal parameters.
Any
The Any component is a Tapestry component that can emulate any type of
markup tag. It has no formal parameters but allows informal parameters. It is
typically used to add simple dynamic output to a normally static tag—for example, to dynamically calculate (using an OGNL expression) the class attribute of a
<span> tag.
Any renders its body and allows informal parameters.
Checkbox
495
Block
A Block component is used to pass a portion of an HTML template (including
static text and components) as a parameter to another component, or even
another page. A RenderBlock component is used to make the Block render its
contents. Chapter 10 includes an example of this, where a portion of a page’s
HTML template was passed into a Palette component to provide part of the Palette’s user interface.
Block renders its body but forbids informal parameters.
Body
The Body component is responsible for the <body> tag of a Tapestry page. As discussed in chapter 8, it is also very much involved in any dynamic JavaScript created by the components within the page.
Body renders its body and allows informal parameters.
Button
The Button component creates an HTML form button element (<input type=
"button"> ). Form buttons are used exclusively to provide client-side actions
(using JavaScript) as part of a form.
Table C.2
Parameters for the Button component
Parameter
name
Type
Direction
Required
Default
value
disabled
boolean
in
No
false
label
string
in
No
Description
If true, the button will be disabled—it
will not respond to the mouse and will
be rendered specially by the browser to
indicate this state.
The label displayed for the button (this
becomes the value attribute of the
output element).
Checkbox
The Checkbox component is one of the standard form element components. It
renders a checkbox (<input type="checkbox">) and reads and updates a boolean
property. The Checkbox component is discussed in detail in chapter 3.
496
APPENDIX C
Table C.3
Parameters for the Checkbox component
Parameter
name
Type
Direction
Required
selected
boolean
in/out
Yes
disabled
boolean
in
No
Default
value
Description
Read when the form is rendered and
updated when the form is submitted.
false
If true, the button will be disabled—it
will not respond to the mouse and will
be rendered specially by the browser to
indicate this state.
Checkbox discards its body and allows informal parameters.
Conditional
The Conditional component is used to render a portion of a page only if a certain condition is met.
The Conditional component normally just renders (or skips) its body. If
desired, the Conditional may also output an element surrounding its body; in
this respect, it operates much as the Any component does.
Table C.4
Parameters for the Conditional component
Parameter
name
Type
Direction
Required
Default
value
Description
condition
boolean
in
No
false
The body of the Conditional is rendered only if this value is true.
invert
boolean
in
No
false
If true, then the condition is
inverted (the body is rendered only
if the condition is false).
element
string
in
No
If given, then an element is output
around the component’s body.
Conditional renders its body and allows informal parameters (which are used
only if the element attribute is specified).
DatePicker
The DatePicker component is a complex form element component that allows
users to select dates using a JavaScript-powered pop-up window.
DirectLink
Table C.5
497
Parameters for the DatePicker component
Parameter
name
Type
Direction
Required
Default
value
Description
value
java.util.Date
in/out
Yes
The date property edited by
the DatePicker component.
format
string
in
No
dd
MMM
yyyy
The date format used to display the date.
disabled
boolean
in
No
false
If true, then the date field
and calendar pop-up buttons
will be disabled.
DatePicker discards its body and forbids informal parameters.
Delegator
The Delegator component is an “escape clause” that allows you to write Java
code to render a portion of a page. You provide the Delegator object with a renderable object (implementing the IRender interface), and it invokes the render()
method on the object. Often, the renderable object is an inner class. An example
of this is in chapter 4, where Delegator was used to produce formatted, hexadecimal output of an uploaded binary file.
Table C.6
Parameter for the Delegator component
Parameter
name
delegate
Type
IRender
Direction
Required
in
Yes
Default
value
Description
The object to invoke to render on
behalf of the Delegator component
Delegator discards its body and forbids informal parameters.
DirectLink
The DirectLink component is the primary way to create hyperlinks in Tapestry
pages; examples of DirectLink appear throughout this book, starting in chapter 2.
DirectLinks allow URLs to be created that reference a particular component
within a page; in addition, a list of parameters may be encoded into the URL,
with the parameter values made available when the link is triggered.
498
APPENDIX C
Table C.7
Parameters for the DirectLink component
Parameter
name
Type
Direction
Required
Default
value
Description
listener
IActionListener
in
Yes
The object (usually a listener
method) to be invoked when
the user triggers the link.
parameters
many
in
No
Zero or more values to be
encoded into the URL and
provided to the listener.
disabled
boolean
in
No
false
If true, the DirectLink
doesn’t render a tag (but
still renders its body).
stateful
boolean
in
No
true
If true (the default), then the
DirectLink will validate that
the HttpSession is still
active when the link is triggered (if a session was active
when the link is rendered).
DirectLink renders its body and accepts informal parameters.
ExceptionDisplay
ExceptionDisplay is a specialized component used by the Tapestry Exception
page to display the stack of exceptions. It is useful if you are creating a customized Exception page.1
Table C.8
Parameter for the ExceptionDisplay component
Parameter
name
exceptions
Type
Exception
Description[]
Direction
in
Required
Yes
Default
value
Description
The stack of exceptions as
generated by the
ExceptionAnalyzer
utility class
ExceptionDisplay discards its body and rejects informal parameters.
1
Note that the portion of the page that displays the Servlet API is produced by using a Delegator, with
the RequestContext as the delegate. RequestContext implements the IRender interface, for just this
purpose.
FieldLabel
499
ExternalLink
The ExternalLink component creates a link to a page, much like the PageLink
component, with two differences: Service parameters may be passed in the URL,
and the target page must implement the IExternalPage interface (the activateExternalPage() method defined by the interface will be invoked). ExternalLink
is used to create links that the end user may safely bookmark, because all the
necessary information is encoded into the URL.
Table C.9
Parameters for the ExternalLink component
Parameter
name
Type
Direction
Required
Default
value
Description
page
string
in
Yes
The name of the page to link to, which
must implement the
IExternalPage interface.
parameters
varied
in
No
Any number of additional parameters
to be encoded into the URL.
disabled
boolean
in
No
false
If true, the ExternalLink doesn’t render
a tag (but still renders its body).
ExternalLink renders its body and accepts informal parameters.
FieldLabel
The FieldLabel component is used to display the label for a ValidField component, integrating into the overall form validation subsystem described in chapter 5.
The FieldLabel coordinates with the enclosing Form’s validation delegate to
visually identify fields within the form that contain errors.
Table C.10
Parameters for the FieldLabel component
Parameter
name
Type
Direction
Required
Default
value
Description
field
form component
in
Yes
The field for which a label
should be displayed
displayName
string
in
No
Overrides the display
name obtained from the
ValidField
FieldLabel renders its body and rejects informal parameters.
500
APPENDIX C
Foreach
The Foreach component is used to iterate over a list of values and repeatedly render its body for each value. It can update a property of its container to identify
the specific value within the list, or the index within the list. A Foreach component can optionally render a tag around its body, much like the Any component.
Table C.11
Parameter
name
Parameters for the Foreach component
Type
Direction
Required
Default
value
null
Description
source
varied
in
No
The list of input values, as a Collection,
Iterator, object array, or single object.
value
Object
out
No
Updated before the Foreach renders its
body.
index
int
out
No
Updated before the Foreach renders its
body.
element
string
in
No
If given, then the Foreach renders a tag
(with informal parameters) around its body
on each pass through the loop.
Foreach renders its body and accepts informal parameters (which are used only if
the element parameter is specified).
Form
The Form component is used to create HTML forms within Tapestry pages, as
discussed in chapters 3 through 5. The Form component is responsible for
orchestrating both the rendering of the form and the processing when the form
is submitted. A Form has a listener that is invoked when the Form is submitted,
after all the enclosed form control elements have had a chance to update properties from the submission.
Table C.12
Parameter
name
listener
Parameters for the Form component
Type
IActionListener
Direction Required
in
No
Default
value
Description
If provided, the listener is
notified after the form
submission is processed.
GenericLink
Table C.12
501
Parameters for the Form component (continued)
Parameter
name
Type
delegate
IValidationDelegate
in
No
stateful
boolean
in
No
Direction Required
Default
value
Description
The validation delegate
used by FieldLabel and
ValidField components
enclosed by the Form.
true
If true and the application
is stateful when the form
is rendered, then a check
occurs that the application is still stateful when
the form is submitted.
The Form component renders its body and accepts informal parameters.
Frame
The Frame component is used within a <frameset> to identify a Tapestry page to
fill a <frame>.
Table C.13
Parameter for the Frame component
Parameter
name
Type
page
string
Direction
in
Required
Default
value
Yes
Description
The name of the page whose content
is rendered within the frame
The Frame component discards its body but accepts informal parameters.
GenericLink
The GenericLink component is used to create hyperlinks to arbitrary URLs. It
may be used to create links to off-site URLs, but it is most often used in conjunction with client-side JavaScript.
Table C.14
Parameters for the GenericLink component
Parameter
name
Type
Direction
Required
href
string
in
Yes
disabled
boolean
in
No
Default
value
Description
The target URL for the link.
false
If true, then the GenericLink does not
render a tag, just its body.
502
APPENDIX C
GenericLink renders its body and accepts informal parameters.
Hidden
The Hidden component is used to record information in an HTML form.
Table C.15
Parameters for the Hidden component
Parameter
name
Type
Direction
Required
Default
value
Description
value
Object
in/out
Yes
The value to be recorded in the
form (and restored when the
form is submitted).
listener
IActionListener
in
No
If provided, the listener is notified after the value parameter is updated when the form
is submitted.
encode
boolean
in
No
true
If true, the value is encoded
with its type. If false, the value
must be a String and is
recorded as is.
The Hidden component discards its body and forbids informal parameters.
Image
The Image component is used to render an image as part of a Tapestry page.
The source of the image is specified as an asset, which hooks into Tapestry’s
localization mechanism to ensure that the correct localization of the image is
automatically used.
Table C.16
Parameters for the Image component
Parameter
name
Type
Direction
Required
image
IAsset
in
Yes
border
int
in
No
Default
value
Description
The image asset to be displayed
0
Sets the width of the border displayed around the image
The Image component discards its body and allows informal parameters.
ImageSubmit
503
ImageSubmit
The ImageSubmit component is closely related to the Submit component; it is
used to display an image that submits a form. When the form is submitted, the
ImageSubmit component can update a property of its container to a specific
value or invoke a listener method (or both). The Submit and ImageSubmit components are discussed in chapter 3.
Table C.17
Parameters for the ImageSubmit component
Parameter name
Type
Direction Required
Default
value
Description
image
IAsset
in
Yes
The image asset to be
displayed.
disabled
boolean
in
No
disabledImage
IAsset
in
No
An optional image to
display instead of the
standard image if the
component is disabled.
point
java.awt.Point
out
No
The point (x and y coordinates) within the
image clicked by the
mouse.
selected
Object
out
No
A property updated by
the ImageSubmit if it is
the cause of the form
submission.
tag
Object
in
No
A value used to update
the selected property
when the ImageSubmit
is the cause of the form
submission.
listener
IActionListener
in
No
A listener invoked by the
ImageSubmit when it is
the cause of the form
submission (after
optionally updating the
selected parameter).
false
If true, the rendered
image will be disabled (it
will not respond to the
mouse).
ImageSubmit discards its body and allows informal parameters.
504
APPENDIX C
Insert
The Insert component is used to insert some text into a rendered Tapestry page.
By default, it escapes out any invalid HTML entities (that is, it converts & to
&, and so on). It may optionally use a java.text.Format object to format an
object before inserting its value.
Table C.18
Parameters for the Insert component
Parameter
name
Type
Direction
Required
Default
value
Description
value
Object
in
No
The value to be inserted.
format
java.text.Format
in
No
If specified, then the value is
formatted to a string before
being inserted.
raw
boolean
in
No
class
string
in
No
false
If true, then no conversion of
HTML entities occurs. This is
used when the value to insert
is known to contain HTML
markup.
If specified, then the output
text is wrapped in a <span>
tag, with a class attribute
set to this value.
The Insert component discards its body and allows informal parameters, which
are used only if the class parameter is specified.
InsertText
The InsertText component is used to insert a block of text into a Tapestry page;
it breaks the text into individual lines and either inserts line breaks between lines
or wraps each line as a separate paragraph. This is commonly used to display
text that was collected using a TextField.
Table C.19
Parameter
name
value
Parameters for the InsertText component
Type
string
Direction Required
in
No
Default value
Description
The multiline text to be
inserted.
LinkSubmit
Table C.19
505
Parameters for the InsertText component (continued)
Parameter
name
Type
mode
InsertTextMode
Direction Required
No
in
Default value
Description
InsertTextMode.BREAK
By default, inserts breaks
between lines. Use
InsertTextMode.
PARAGRAPH to wrap each
line as a paragraph.
InsertText discards its body and forbids informal parameters.
LinkSubmit
The LinkSubmit component is used when you wish to have a hyperlink on a
page cause a form to submit. This is useful when using a Rollover component
inside the link. LinkSubmit requires client-side JavaScript.
Table C.20
Parameters for the LinkSubmit component
Parameter
name
Type
Direction
Required
Default
value
false
Description
disabled
boolean
in
No
If true, the LinkSubmit will
not render a tag, but will still
render its body.
selected
Object
out
No
A property updated by the
LinkSubmit when it is the
cause of the form submission.
tag
Object
in
No
A value used to update the
selected property when
the LinkSubmit is the cause
of the form submission.
listener
IActionListener
in
No
LinkSubmit when it is the
cause of the form submission (after optionally updating the selected
parameter).
LinkSubmit renders its body and allows informal parameters.
506
APPENDIX C
ListEdit
The ListEdit component is used within a Form to iterate over a list of values in a
fashion that is compatible with use within a Form. The ListEdit component
records into the Form the values it iterates over during the render (as a number
of hidden fields). This ensures that the form submission is processed consistently
with the Form’s render. ListEdit was covered in detail in chapter 4.
Table C.21
Parameters for the ListEdit component
Parameter
name
Type
Direction
Required
Default
value
null
Description
source
varied
in
No
The list of values to iterate
over, as a List, Iterator,
or object array.
value
Object
out
No
Set by the ListEdit on each
pass through the loop.
index
int
out
No
Set by the ListEdit on each
listener
IActionListener
in
No
Invoked by the ListEdit just
after the value parameter is
set (both when rendering and
when rewinding the form).
element
string
in
No
If specified, the ListEdit writes
a tag around its body on each
ListEdit renders its body and accepts informal parameters.
Option
The Option component renders an HTML <option> tag. Option components
must be enclosed by a Select component. In most cases, it is easier to use a PropertySelection component than to build the equivalent using Select and Option.
Select and Option were covered in chapter 3.
Table C.22
Parameters for the Option component
Parameter
name
Type
Direction
selected
boolean
in/out
Required
Yes
Default
value
Description
Set to true if the option is selected when
the form is submitted
PropertySelection
Table C.22
507
Parameters for the Option component (continued)
Parameter
name
Type
label
Direction
string
Required
Default
value
No
in
Description
A string used as the label of the option
Option renders its body (which is another way to specify its label) and accepts
informal parameters.
PageLink
The PageLink component is used to create a link to a specific Tapestry page. The
indicated page is rendered as the response when the link is clicked.
Table C.23
Parameters for the PageLink component
Parameter
name
Type
Direction
Required
page
string
in
Yes
disabled
boolean
in
No
Default
value
Description
The name of the page to be rendered.
false
If true, the PageLink component
does not render its tag but still renders its body.
PageLink renders its body and accepts informal parameters.
PropertySelection
The PropertySelection component is used to create a drop-down list for selecting
a value for a single property. As discussed in chapter 4, PropertySelection uses a
model to provide the possible values and the labels for those values.
Table C.24
Parameters for the PropertySelection component
Parameter
name
Type
Direction
Required
Default
value
Description
value
Object
in/out
Yes
The property to edit; the
property is read during render and updated when the
form is submitted.
model
IProperty
SelectionModel
in
Yes
Used to identify the possible
values and associated labels.
508
APPENDIX C
Table C.24
Parameters for the PropertySelection component (continued)
Parameter
name
Type
Direction
Required
Default
value
Description
disabled
boolean
in
No
false
If true, the PropertySelection is inactive.
submitOnChange
boolean
in
No
false
If true, changing the value
will cause the enclosing
form to submit.
PropertySelection discards its body and allows informal parameters.
Radio
The Radio component creates a radio button within a form. Radio components
must be enclosed by a RadioGroup. Radio and RadioGroup were discussed in
chapter 3.
Table C.25
Parameters for the Radio component
Parameter
name
Type
Direction
Required
value
Object
in
No
disabled
boolean
in
No
Default
value
Description
The value is used to determine if the
radio button is initially selected (when
rendering) and is the value assigned to
the selected parameter when the
form is submitted, if the HTML radio
button is selected.
false
If true, the radio button will be disabled
and will not respond to the mouse.
Radio discards its body and allows informal parameters.
RadioGroup
The RadioGroup component acts like a container for any number of Radio components. It is responsible for identifying the selected Radio component and
updating a property to match the selection. Radio and RadioGroup were discussed in chapter 3.
Rollover
Table C.26
509
Parameters for the RadioGroup component
Parameter
name
Type
Direction
Required
selected
Object
in/out
Yes
disabled
boolean
in
No
Default
value
Description
The property updated when the form
is submitted.
false
If true, all Radio components within
the group are disabled.
RadioGroup renders its body but rejects informal parameters.
RenderBlock
The RenderBlock component is used to make a Block component renders its
body. In effect, a portion of a page may be passed as a parameter to a component. Block and RenderBlock are commonly used to create a tabbed view.
Table C.27
Parameters for the RenderBlock component
Parameter
name
block
Type
Block
Direction
in
Required
No
Default
value
Description
The Block whose body should be
rendered
Block discards its body and allows informal parameters.
RenderBody
The RenderBody component allows a component to insert its body at a prescribed position within its own template. RenderBody takes no parameters and
discards its body.
Rollover
The Rollover component is a specialized version of the Image component that is
used within some form of link component (DirectLink, ExternalLink, GenericLink, and so on). Rollover can perform mouse rollovers (displaying a different
image when the user moves the mouse over the image). In addition, a Rollover
component can adjust the image displayed when the enclosing link is disabled.
510
APPENDIX C
Table C.28
Parameters for the Rollover component
Parameter
name
Type
Direction
Required
Default
value
Description
image
IAsset
in
No
The initial image to display.
focus
IAsset
in
No
If given, the image to display when
the mouse is moved over the image.
blur
IAsset
in
No
If given, the image to display when
the mouse is moved out from over
the image. If not specified, the normal image is displayed.
disabled
IAsset
in
No
If given, the image to display instead
of the default image if the enclosing
link is itself disabled.
Rollover discards its body and allows informal parameters.
Script
The Script component is used to construct client-side JavaScript from a script
specification. Informal parameters to the Script are added as additional input
symbols. Dynamic script creation was discussed in chapter 8. Script components
must be enclosed by a Body component.
Table C.29
Parameter
name
Parameters for the Script component
Type
Direction
Required
Default
value
Description
script
string
in
Yes
The name of the script specification file
symbols
Map
in
No
If given, provides the base set of input
symbols for the script
Script discards its body and allows informal parameters.
Select
A Select component renders an HTML <select> element and encloses a number
of Option components. In most cases, using a PropertySelection component
achieves the same result and is easier; the Select component is most often used
when multiple selection is desired. An example of using Select and Option for
multiple selection appears in chapter 3.
Shell
Table C.30
511
Parameters for the Select component
Parameter
name
Type
Direction
Required
Default
value
Description
multiple
boolean
in
No
false
If true, then multiple selection will
be allowed.
disabled
boolean
in
No
false
If true, the rendered <select> will
include the disabled attribute.
Select renders its body and allows informal parameters.
ServiceLink
The ServiceLink component is used to render a hyperlink that references a
named engine service. This is often used to invoke the restart engine service, but
may also be used with custom application services.
Table C.31
Parameters for the ServiceLink component
Parameter
name
Type
Direction
Required
Default
value
Description
service
string
in
Yes
The name of the engine service.
parameters
varied
in
No
Objects to be encoded into the URL
(as with DirectLink).
disabled
boolean
in
No
false
If true, the ServiceLink does not render its tag but still renders its body.
ServiceLink renders its body and accepts informal parameters.
Shell
The Shell component is a convenience for rendering the <html> and <head> portions
of a page. It is particularly useful for adding one or more stylesheets to a page.
Table C.32
Parameters for the Shell component
Parameter
name
Type
Direction
Required
Default value
Description
title
string
in
Yes
The title of the page
stylesheet
IAsset
in
No
A single stylesheet to
include
512
APPENDIX C
Table C.32
Parameters for the Shell component (continued)
Parameter
name
Type
Direction
Required
Default value
Description
stylesheets
Array or
collection of
IAsset
in
No
A list of stylesheets to
include
refresh
int
in
No
If specified, the time
(in seconds) before
redisplaying the same
page
DTD
string
in
No
delegate
IRender
in
No
-//W3C//DTD
HTML 4.0
Transitional//EN
Used to specifying the
<!DOCTYPE> of the
page
A renderer, used to
add content to the
<head> element
The Shell component renders its body and forbids informal parameters.
Submit
The Submit component renders a submit button within a form. When the form is
submitted due to the rendered button, the Submit component can update a
property or invoke a listener method (or both).
Table C.33
Parameters for the Submit component
Parameter
name
Type
Direction
Required
Default
value
Description
selected
Object
out
No
A property updated by the
Submit if it is the cause of
the form submission.
tag
Object
in
No
A value used to update the
selected property when
the Submit is the cause of
the form submission.
listener
IActionListener
in
No
Submit when it is the cause
of the form submission
(after optionally updating
the selected parameter).
TextField
Table C.33
513
Parameters for the Submit component (continued)
Parameter
name
disabled
Type
Direction
boolean
Required
No
in
Default
value
false
Description
If true, the button will be
disabled (it will not respond
to the mouse).
Submit discards its body and accepts informal parameters.
TextArea
The TextArea component creates a multiline input field as an HTML <textarea>
element.
Table C.34
Parameters for the TextArea component
Parameter
name
Type
Direction
Required
value
string
in/out
No
disabled
boolean
in
No
Default
value
Description
The value to edit, which is read when
the TextArea renders and updated
when the enclosing form is submitted.
false
If true, the TextArea is disabled and
will not update the value parameter.
TextArea discards its body and accepts informal parameters.
TextField
The TextField component is used to create an HTML text or password field
within a form. TextField is described in detail in chapter 3.
Table C.35
Parameters for the TextField component
Parameter
name
Type
Direction
Required
value
string
in/out
No
hidden
boolean
in
No
Default
value
Description
The value to edit, which is read when
the TextField renders and updated
when the enclosing form is submitted.
false
If true, the field is rendered as a
password field (which hides the exact
text entered by the user).
514
APPENDIX C
Table C.35
Parameters for the TextField component (continued)
Parameter
name
disabled
Type
boolean
Direction
Required
No
in
Default
value
false
Description
If true, the field is rendered with a
disabled attribute, and the component does not update the value
parameter when the form is submitted.
TextField discards its body and accepts informal parameters.
Upload
The Upload component allows users to upload files from their computer to the
server. Upload is a form component and must be enclosed by a Form. Upload is
discussed in detail in chapter 4.
Table C.36
Parameters for the Upload component
Parameter
name
Type
Direction
Required
file
IUploadFile
out
Yes
disabled
boolean
in
No
Default
value
Description
The uploaded file is represented
by an instance of
IUploadFile, assigned to this
parameter.
false
If true, the HTML <input> element will include a disabled
attribute.
Upload discards its body and allows informal parameters.
ValidField
The ValidField component is a specialized version of TextField that is integrated into the Tapestry validation subsystem, described in chapter 5. A ValidField is not limited to editing string properties (as TextField is) but can edit
any data type, as long as it has an appropriate validator to perform the necessary conversions.
ValidField
Table C.37
515
Parameters for the ValidField component
Parameter
name
Type
Direction
Required
Default
value
Description
value
Object
in/out
Yes
The property to be edited by the
ValidField.
validator
IValidator
in
Yes
An object that validates user
input and translates between
strings and object values.
displayName
string
in
Yes
The name of the field, used in
error messages and by the FieldLabel component.
hidden
boolean
in
No
false
If true, the component renders a
password field, not a normal text
input field.
disabled
boolean
in
No
false
If true, the <input> tag will
include a disabled attribute,
and the ValidField will not update
the value parameter when the
form is submitted.
ValidField discards its body and accepts informal parameters.
516
Primary specifications
517
Tapestry applications include five types of specifications: page, component, application, library, and script. All of these specifications are XML documents, validated
against a document type definition (DTD). The first four specifications are the primary specifications, and they share a single DTD (but use different document root
elements). The script specification uses its own, separate DTD.
D.1 Primary specifications
The primary specifications use the following XML <!DOCTYPE>:
<!DOCTYPE root element PUBLIC
Each type of specification has its own file extension and a specific root element,
shown in table D.1.
Table D.1
Tapestry specifications and file extensions
Type
File extension
Root element
application
.application
<application>
page
.page
<page-specification>
component
.jwc
<component-specification>
library
.library
The following sections identify each element, explaining how it’s used and listing its attributes and nested elements.
<application>
The <application> element is the root element of an application specification. Application specifications are generally optional and are used in only a few situations:
■
When a page or component specification is not stored in an expected location
■
When the application defines custom engine services
■
When the application makes use of a component library (as discussed in
chapter 6)
■
When the application defines any application extensions (as discussed in
chapter 8)
Figure D.1 identifies the elements that may be contained within the <application> element. Table D.2 lists the attributes that may be specified.
518
APPENDIX D
Figure D.1 The <application> element is primarily used to override the default rules for locating
pages and components, and to specify libraries, extensions, and services.
NOTE
Table D.2
These diagrams express the rules for the content of an element. Elements contain a sequence of other elements, with a symbol to indicate
how many of that element are allowed (? indicates optional, * indicates
zero or more, + indicates at least one). Choice of means that exactly
one of the listed elements is allowed. Figure D.1 indicates that an <application> can contain an optional <description>, followed by any
number of <property> elements, followed by any number of the remaining elements.
Attributes of the <application> element
Attribute name
Required?
Description
engine-class
No
Name of the class to instantiate as the application engine; if not
specified, BaseEngine is used.
name
No
A user-presentable name for the application, used in some debugging
output.
<bean>
The <bean> element is used to create a helper bean, a Java bean that provides
additional logic to a page. <bean> appears inside <page-specification> and
<component-specification>.
Frequently, <bean> is used to create the validation delegate and individual validator objects used with the form validation subsystem described in chapter 5.
Figure D.2 identifies the elements that may be contained by <bean>, and table D.3
lists the attributes for the <bean> element.
Figure D.2
Table D.3
519
The <bean> element is used to define a helper bean and configure its properties.
Attributes of the <bean> element
Attribute name
Required?
Description
class
Yes
The Java class to instantiate.
lifecycle
No
The lifecycle of the bean: none, request, page, or render. The default
is request.
name
Yes
The name of the bean, which must be a valid Java identifier.
<binding>
The <binding> element appears inside the <component> element. It binds a
parameter of the component to an OGNL expression. The expression, which is
most often simply the name of a property, is evaluated in the context of the containing page or component.
OGNL expressions can be very complex and involve characters, such as single
and double quotes, that are awkward to express within an XML attribute. To
accommodate such situations, the OGNL expression can be provided as the
wrapped character data enclosed by the <binding> element.
Table D.4 lists the attributes for the <binding> element. <binding> may not
contain any other elements.
Table D.4
Attributes of the <binding> element
Attribute name
Required?
Description
expression
No
The OGNL expression. If not specified, the content of the element is
used as the expression.
name
Yes
The name of the component parameter to bind; this may be either a
formal or informal parameter.
520
APPENDIX D
<configure>
The <configure> element is used within the <extension> element to configure
one property of the application extension. <configure> may not contain other
elements. Table D.5 lists the attributes of the <configure> element.
Table D.5
Attributes of the <configure> element
Attribute name
Required?
Description
property-name
Yes
The name of the property of the extension to configure.
type
No
The type of the property, which is used to convert the value:
boolean, int, long, double, or String. String is the default.
value
No
The value to be assigned. If not provided as an attribute, the character data enclosed by the element is used as the value.
<component>
The <component> element declares the use of a component within another page
or component. It identifies the ID of the component, its type, and the configuration of any of its formal and informal parameters. Figure D.3 identifies the elements that may be enclosed by <component>; these are mostly different elements
for binding parameters of the component.
In most cases, the <component> element defines the type of component. In
some cases, it is convenient to copy another component’s type and bindings, and
optionally extend or override the bindings. This can be accomplished using the
copy-of attribute instead of the type attribute. Table D.6 lists the attributes of the
<component> element.
Figure D.3 The <component> element primarily contains different types of bindings for the
parameters of the component.
Table D.6
521
Attributes of the <component> element
Attribute name
Required?
Description
copy-of
No
Identifies another component, already defined in the containing page
or component, to use as a template for this component. May not be
specified with the type attribute.
id
Yes
The ID used for the component, which must be unique within the container and must be a valid Java identifier.
inheritinformalparameters
No
If yes, then the component inherits informal parameters from its container. The default is no.
type
No
The type of component to create. The type may include a prefix to
identify the library that contains the component. May not be specified
with the copy-of attribute.
<component-type>
The <component-type> element is used within an application specification
(<application>) or library specification (<library-specification>) to define a
type of component. Normally, this is not needed; in most cases, putting the component specification (the .jwc file) in the same folder as the application or library
specification is sufficient. You use <component-type> when the specification file is
not stored in the default location.
Keep in mind that <component-type> may not contain other elements. Table D.7
lists the attributes for <component-type>.
Table D.7
Attributes for the <component-type> element
Attribute name
Required?
Description
specification-path
Yes
The location, relative to the application or library specification, of the component’s specification file
type
Yes
The component type to define
<component-specification>
The <component-specification> element is the root element of a component
specification file (with a .jwc extension). A component specification defines a
reusable component, including its formal parameters, and specifies whether it
uses or discards its body, as well as embedded components and assets. Figure D.4
shows the elements that may be enclosed by <component-specification>.
The <component-specification> element’s class attribute is used to identify
the Java class for the component. If the component has formal parameters or
522
APPENDIX D
Figure D.4 The <component-specification> element defines the parameters, beans,
components, and assets used by the component.
any specified properties, then Tapestry will create a subclass to provide the attributes
and accessor methods for those properties. Table D.8 lists the attributes for the
<component-specification> element.
Table D.8
Attributes for the <component-specification> element
Attribute name
Required?
Description
allow-body
No
If yes (the default), then the component keeps its body; if no, the
component discards its body (and the body may not define components).
allowinformalparameters
No
If Yes (the default), then the component will allow informal parameters. If no, then informal parameters are not allowed.
class
No
The Java class for the component, or BaseComponent if not specified. Tapestry will create a subclass to contain specified properties
and parameters.
523
<context-asset>
The <context-asset> element is one of three elements used to define assets.
(Assets are described in chapters 2 and 6.) Use <context-asset> to define an
asset stored within your web application context. The element may appear inside
<page-specification> or <component-specification>. The asset path is relative
to the web application context root. The <context-asset> element may contain
the <property> element. Table D.9 lists its attributes.
Table D.9
Attributes for the <context-asset> element
Attribute name
Required?
Description
name
Yes
The name of the asset to define, which must be a valid Java identifier
path
Yes
The path of the asset, relative to the context root
<description>
The <description> element appears inside many other elements and allows a
user-presentable description to be attached to many elements. It is always
optional, and the framework does not directly use the description. The descriptive text is the character content enclosed by the <description> element. This
element may not enclose any other elements.
<extension>
The <extension> element is used to create an application extension. Application
extensions are similar in intent to helper beans (the <bean> element) but are global to the entire application. The <extension> element appears inside the
<application> element (of an application specification), or inside the <libraryspecification> element of a library specification. Figure D.5 shows the elements
that may be contained within an <extension> element.
The <extension> element defines a JavaBean to instantiate. Normally, an
extension is instantiated only when first referenced (see the Javadoc for the
ILibrarySpecification interface). The immediate attribute is used to force the
extension to be instantiated when the application initializes. Table D.10 lists the
attributes of the <extension> element.
Figure D.5
The <extension> element
contains metadata (<property>)
and <configure> elements to
set the properties of the extension.
524
APPENDIX D
Table D.10
Attributes of the <extension> element
Attribute name
Required?
Description
class
Yes
The name of the Java class to instantiate.
immediate
No
If yes, then the extension is instantiated and configured at startup; if
no (the default), then the extension is instantiated lazily, on first reference.
name
Yes
The name of the extension, which is usually qualified with a Java
package prefix (i.e., org.apache.tapestry.property-source).
<external-asset>
The <external-asset> element is one of three used to define assets. (Assets are
described in chapters 2 and 6.) The <external-asset> element is used to define
an asset at an arbitrary URL. External assets are never localized; they are always
used as is. This element may appear inside <page-specification> or <componentspecification>, and it may contain the <property> element. Its attributes are
listed in table D.11.
Table D.11
Attributes of the <external-asset> element
Attribute name
Required?
Description
name
Yes
URL
Yes
The URL of the asset, which is used as is
<inherited-binding>
The <inherited-binding> element is used to bind a parameter of a component
to a parameter of its container. This element appears inside the <component> element and may not contain other elements. Table D.12 lists the attributes for the
<inherited-binding> element.
Table D.12
Attributes of the <inherited-binding> element
Attribute name
Required?
Description
name
Yes
The name of the component parameter to bind
parameter-name
Yes
The name of the containing component’s parameter to bind
525
<library>
The <library> element is used in an application specification (<application>) or
library specification (<library-specification>) to identify a library that may be
referenced in the page and component specifications of the application or referencing library. The name provided becomes a prefix on the component type.
This is demonstrated in chapters 7 and 8, where the Palette component (from
the Tapestry contributions library) is used.
The <library> element may not contain other elements. Its attributes are
listed in table D.13.
Table D.13
Attributes of the <library> element
Attribute name
Required?
Description
id
Yes
The ID for the library, which must be a valid Java identifier. The ID is
used as a prefix to access the components within the library.
specificationpath
Yes
The path to the library. If a relative path, then it is relative to the referencing application or library specification; if a complete path
(starting with a /), then the specification is expected to be stored
on the classpath (i.e., within a JAR file).
The <library-specification> element is the root element of a library specification. It is similar to the <application> element, except that library specifications
are not optional—a library must have a specification, even if it is empty.
The <library-specification> element can contain the same elements as
<application> (see figure D.1 for a listing). The <library-specification> element does not have any attributes.
<listener-binding>
The <listener-binding> element allows listeners (for Form, DirectLink, and
similar components) to be specified, in the page or component specification,
using a scripting language such as Jython. This functionality is targeted at rapid
application development and is still experimental. The character content
enclosed by the <listener-binding> element consists of the script you want to
execute. Table D.14 lists the attributes for the <listener-binding> element.
The <listener-binding> element is based on the open source Bean Scripting
Framework (http://jakarta.apache.org/bsf), a runtime environment that supports
multiple scripting languages. The default scripting language is Jython (Python
526
APPENDIX D
Table D.14
Attributes for the <listener-binding> element
Attribute name
Required?
Description
language
No
The name of the scripting language used by the script
name
Yes
The name of the parameter of the component to bind to the listener
implemented in Java). More information on Jython is available from
www.jython.org/. Jython is not distributed with Tapestry, and you must download
it separately and include it in the runtime classpath.
<message-binding>
The <message-binding> element appears inside a <component> element. It is used
to bind a parameter of a component to a literal value from the containing component’s localized message properties. The <message-binding> element may not
contain other elements. Table D.15 lists the attributes of this element.
Table D.15
Attributes for the <message-binding> element
Attribute name
Required?
Description
name
Yes
The name of the component parameter to bind
key
Yes
The key used to obtain the localized message
<page>
The <page> element is used to define a page, much like a <component-type> defines
a component. It appears within the <application> and <library-specification>
elements, and is only necessary when the page specification cannot be located by
the normal search path.
The <page> element may not enclose other elements. The attributes for the
<page> elements are shown in table D.16.
Table D.16
Attributes for the <page> element
Attribute name
Required?
Description
name
Yes
The name of the page, which must be a valid Java identifier
specificationpath
Yes
The location of the page specification, relative to the application or
library specification
527
Figure D.6 The <page-specification> element is similar to the <componentspecification> element, with elements related to parameters omitted.
<page-specification>
The <page-specification> element is the root element for a Tapestry page specification. Page specifications are similar to component specifications, with those
attributes and elements related to parameters removed (pages do not have
parameters). The common elements are those that describe embedded components, assets, beans, and properties.
Figure D.6 identifies the elements that may be enclosed within the <pagespecification> element.
Table D.17 lists the lone attribute for the <page-specification> element,
class. When the <property-specification> element is used to define new properties, Tapestry will create a subclass containing the necessary attributes and
accessor methods.
Table D.17
Attribute for the <page-specification> element
Attribute name
class
Required?
no
Description
The page’s class, or BasePage if not specified
<parameter>
The <parameter> element is used to define a formal parameter inside a <componentspecification>. The <parameter> element may contain a <description> element,
528
APPENDIX D
which is used to document the parameter. Table D.18 lists the attributes of the
<parameter> element. Components and component parameters are discussed in
chapter 6.
Table D.18
Attributes for the <parameter> element
Attribute
name
Required?
Description
defaultvalue
No
A default value for the parameter, if not specified, as an OGNL expression.
direction
No
The direction of the parameter, as discussed in chapter 7. The attribute
value can be in, form, auto or the default, custom.
name
Yes
The name of the parameter, which must be a valid Java identifier.
propertyname
No
If specified, then the property for the parameter will be created with the
given name rather than the parameter name. This is used when the
parameter name collides with an existing parameter, such as the page
parameter on the PageLink component.
required
No
If yes, then the parameter must be bound. If no (the default), then it is
acceptable to omit the parameter.
type
No
The name of a Java class or primitive type; this is used when creating a
property for the parameter. If not specified, then java.lang.Object is
used.
<private-asset>
The <private-asset> is one of three elements used to define assets. (Assets are
described in chapters 2 and 6.) The <private-asset> element is used to define
an asset stored on the classpath. Typically, such a resource is distributed as part
of a JAR containing a component library. The <private-asset> element may
appear inside <page-specification> or <component-specification> . The
resource path is relative to the specification file. The <private-asset> element
may contain the <property> element; its attributes are listed in table D.19.
Table D.19
Attributes for the <private-asset> element
Attribute name
Required?
Description
name
Yes
resource-path
Yes
The path to the asset
529
<property>
The <property> element is used to define metadata about the element which contains it;1 it is contained by many elements, notably <application>, <componentspecification>, and <page-specification>. Tapestry makes only limited use of
the <property> element (such as to define the Visit object’s class). Chapter 10
includes an example of an application-specific use of the <property> element.
Beyond that, this element is provided as a flexible hook to be exploited by future
Tapestry tools; for example, it could be used by a Model-Driven Architecture (MDA)
tool to facilitate round-trip engineering of a generated Tapestry application.
In any event, the <property> element contains no other elements. Table D.20
lists the attributes of this element.
Table D.20
Attributes for the <property> element
Attribute name
Required?
Description
name
Yes
The name of the metadata property.
value
No
The value of the property; if not specified, then the enclosed character
data is the value for the property.
<property-specification>
The <property-specification> element directs Tapestry to create a new property within the page or component class. Tapestry will create a subclass, adding
the necessary attributes and accessor methods (as well as any notifications that
are needed).
An OGNL expression may be provided as the initial value for the property;
this can be specified as the initial-value attribute or as the enclosed character
data. The property may be either persistent or transient. Persistent properties
are restored at the start of each request cycle, whereas transient properties are
reset to the initial value at the end of each request cycle. Table D.21 lists the
attributes for the <property-specification> element. This element may not
enclose other elements.
1
In hindsight, the name <meta> may have been more appropriate.
530
APPENDIX D
Table D.21
Attributes for the <property-specification> element
Attribute name
Required?
Description
initialvalue
No
An OGNL expression used to set the initial value for the property. If
omitted, the character data enclosed by the element is used.
name
Yes
The name of the property to create.
persistent
No
If yes, then the value for the property will persist between request
cycles. If no (the default), the value will be reset to the initial value at
the end of each request.
type
No
The type of the property: a Java class name or primitive type. If omitted, the default is java.lang.Object.
<reserved-parameter>
The <reserved-parameter> element is enclosed by the <component-specification>
element; it is only used with components that allow informal parameters. Each
name specified by a reserved parameter is not allowed as an informal parameter.
This is used by components that emit their own HTML attributes to indicate
which HTML attributes are reserved to the component. For example, the DirectLink component reserves the href parameter, since it generates that parameter
in its own implementation.
Each informal parameter is matched, ignoring case, against all reserved
names. Informal parameters that match a reserved name or a formal parameter
name are discarded.
Table D.22 identifies the sole attribute of the <reserved-parameter> element;
<reserved-parameter> may not contain other elements.
Table D.22
Attribute for the <reserved-parameter> element
Attribute name
name
Required?
Yes
Description
A name to reserve
<service>
The <service> element is used to define a new engine service (as described in
chapters 7 and 8). This element may be contained within an application specification (<application>) or a library specification (<library-specification>).
Tapestry starts with a basic set of services; it is possible to override the default
implementations of these services or create new services. New services should be
given names that avoid conflicts, typically by prefixing the name with a Java
package name.
531
The <service> element may not contain other elements. Table D.23 lists
its attributes.
Table D.23
Attributes for the <service> element
Attribute name
Required?
Description
class
Yes
The name of the Java class to instantiate as the service. The class
must implement the IEngineService interface.
name
Yes
The name of the service.
<set-property>
The <set-property> element is contained within a <bean> element; it is used to
set a property of the instantiated bean to a value computed from an OGNL
expression. The <set-property> element may not contain other elements.
Table D.24 lists the attributes for this element.
Table D.24
Attributes for the <set-property> element
Attribute name
Required?
Description
expression
No
The OGNL expression to set the property to. If not specified, the
enclosed content of the element is the source of the expression.
name
Yes
The name of the property to set.
<static-binding>
The <static-binding> element appears inside the <component> element, and it
is used to bind a parameter of a component to a static (or literal) value. It may
not enclose other elements. Table D.25 lists the attributes for the <staticbinding> element.
Table D.25
Attributes for the <static-binding> element
Attribute name
Required?
Description
name
Yes
The name of the component parameter to bind.
value
No
The literal value to bind the parameter to. If not specified, the
enclosed character content of the element is the literal value.
532
APPENDIX D
D.2 Tapestry script specification
Tapestry script specifications are used when generating dynamic JavaScript for
a Tapestry page, and are covered in chapter 8. Script specifications consist of
five sections:
■
Including static JavaScript libraries
■
Defining input symbols
■
Creating new symbols by combining the input symbols
■
Producing the script body
■
Producing the script initialization
Figure D.7 shows this structure reflected in the construction of the <script> element, the root element of a script specification.
The <!DOCTYPE> for a script specification must be
3.0//EN"
➥
Figure D.7
The <script> element is the root element of a Tapestry script specification.
Tapestry script specification
Figure D.8
533
The <body> element contains text, OGNL expressions, and the other main elements.
<body>
The <body> element makes a contribution to the page’s main JavaScript block.
This element, which takes no attributes, contains a mix of text and elements
shown in figure D.8.
Static text is contributed to the JavaScript block as is. Text may also contain
expressions, using the Ant-like syntax ${expression}. See chapter 8 for examples
of this.
<foreach>
The <foreach> element is analogous to the Foreach component; it iterates over a
list of items, rendering its body repeatedly. The <foreach> element may contain
the exact same content as the <body> component (shown in figure D.8).
Like the Foreach component, the <foreach> element has attributes for defining the input list, for setting the current item from the list, and for setting the
index within the list. These attributes are shown in table D.26.
Table D.26
Attributes of the <foreach> element
Attribute name
Required?
Description
expression
Yes
The OGNL expression is evaluated to provide the list of items to iterate
over; this may be a single object, an array of objects, or a List of
objects.
index
No
If provided, then the index within the loop is stored as a symbol with
the provided key.
key
Yes
The key used to store the value of the current item from the list.
534
APPENDIX D
<if>
The <if> element evaluates an expression and renders its body only if the
expression evaluates to true. This element may contain the exact same content as
the <body> component (shown in figure D.8). Table D.27 lists the single attribute
of the <if> element.
Table D.27
Attribute of the <if> element
Attribute name
Required?
expression
Yes
Description
The OGNL expression to evaluate
<if-not>
The <if-not> element is exactly the same as the <if> element, with one exception: It renders its body only if the expression evaluates to false.
<include-script>
The <include-script> element is used to include a common JavaScript library as
part of the page. Tapestry will ensure that the library will be included only once,
regardless of how many times a script specification is executed (or how many
script specifications include the same script). This may be useful when you’re
integrating a third-party JavaScript library into a Tapestry application.
The JavaScript library must be stored on the classpath, typically as part of a
JAR file containing a component library. The path can be specified as a relative
path, in which case the path is relative to the script specification (this should be
used only with script specifications that are stored on the classpath). If the path is
absolute (starting with a leading /), then the location is always assumed to be on
the classpath, regardless of the location of the script specification.
Table D.28 lists the attribute of the <include-script> element.
Table D.28
Attribute of the <include-script> element
Attribute name
resource-path
Required?
Yes
Description
The path to the library
<initialization>
The <initialization> element is exactly the same as the <body> element, with
one exception: The content of the <initialization> element is placed in a JavaScript block that executes only when the page loads. The <initialization> content
Tapestry script specification
535
is typically used to connect client-side objects to event handlers (with the event
handlers defined within the <body> element).
<input-symbol>
The <input-symbol> element is used to define an input symbol to a script specification. Defining input symbols is optional but useful; it allows the framework to
verify that all required input symbols are specified and that the values passed are
of the correct type.
The <input-symbol> element may not contain other elements; its attributes
are defined in table D.29.
Table D.29
Attributes of the <input-symbol> element
Attribute name
Required?
Description
class
No
The Java class (or interface) that the symbol value must be assignable
to; if not specified, java.lang.Object is used.
key
Yes
The input symbol to be defined.
required
No
If yes, then a non-null value must be supplied to the symbol. If no
(the default), then it is acceptable to omit the symbol.
<let>
The <let> element is used to create a new symbol by combining existing symbols. The body of the <let> element is evaluated and converted into a string,
which is trimmed of leading and trailing whitespace and assigned as a new symbol. The symbol may be further modified to ensure that it is unique throughout
the entire page (as discussed in chapter 8).
The <let> element contains the same content as the <body> element (in figure D.8). Table D.30 lists the attributes of the <let> element.
Table D.30
Attributes of the <let> element
Attribute name
Required?
Description
key
Yes
The key for the new symbol.
unique
No
If yes, then the value may be modified to ensure that it is unique.
<set>
The <set> element is used to create a new symbol by evaluating an OGNL expression. The <set> element may not contain any content. The new symbol is not
536
APPENDIX D
necessarily a text value; it is whatever type results from evaluating the expression. Table D.31 lists the attributes for the <set> element.
Table D.31
Attributes of the <set> element
Attribute name
Required?
Description
expression
Yes
The OGNL expression to evaluate
key
Yes
The key for the new symbol
<unique>
The <unique> element defines an area within the specification that is evaluated
only once per render of the page, regardless of how many times the script specification is executed. This is useful for producing portions of the page’s JavaScript block that contains invariant sections. This is similar to including the
content of an external JavaScript library (as with the <include-script> element)
but doing so inline.
The <unique> element takes no attributes and allows the same content as the
<body> element (shown in figure D.8).
index
Symbols
@ 55, 84
$content$ 245
$remove$ 83, 184
/app 90
A
absolute paths 371
abstract methods 99–101
AbstractComponent 33, 59, 75
AbstractPage 147
accessor methods 81, 101, 128,
223, 225, 228, 237, 308, 529
action service 494
Actions 37
activate 62
ActivateCallback 445
ActivatePage 411, 444
active page 62, 100
AdminPage 411
advertising sites, internationalization of 309
Ant 29
Ant build scripts 385
anti-patterns 13, 15
bad coding shortcuts 15
unwanted dependencies 13,
98
weak binding 13, 95
Apache 25, 480
Apache Tomcat 7
application callback 55
application extensions 328, 520
application flow 41
application layer 19–20, 404,
411
application server cluster 409
application servers 11, 18
class loaders 273
clusters 275
hot application redeployment 273
JBoss 405, 480
servlets
See application servlets
WebLogic 276, 405
application servlets
application operations, handling 280
as gateways 271–272
bridging between HTTP and
299
clusters and 272
engine services and 279
HTTP protocol 271
JavaBeans 271
mapping changes 279
multithreading, impact on
271
overview 271
page use compared to use of
300
relationship between JSPs and
279
request processing 270–271
return of engine instance to
HttpSession 272
Servlet API 271, 273
stale session 285
stateful 285
stateless 272
URLs 279, 284, 290
application specifications 91,
517
application tier 44
ApplicationRuntimeException
53
applications
banner source in specifications 339
border components and 217
changing locales for 317
compartmentalizing 262
continuing operation with
missing message strings
315
Demo 12
desktop 13, 16
engines and 277
enterprise 12
examples of
See example applications
extremely large, Tapestry
framework and 274
file extension of specifications
254
forwarding requests to 369
highly complex, Tapestry
framework and 274
hybrid 369
integrating with JSP 369
JavaBean construction of 299
language proficiency for 263
libraries and 256
library for services and components 336
537
538
INDEX
applications (continued)
linking from JSPs to 371
localization of 309
look and feel of 217
multi-user, Tapestry framework and 274
overriding default locations
of component specifications 256
resource storage 261
root element for specifications 256
servlets
See application servlets
specifications 254
stateless 61
URLs 333, 371
usable 170, 192, 211
user expectations 170
user-presentable names for
256
validation of input 179
web, localization of 309
ApplicationServlet 276, 409
application-specific logic 5, 22,
57
architecture 16
array types 228
ArrayList 126
asset services 267
assets 70, 75–76, 79, 184, 437,
523
context 267
duplication removal 244
image representation with
316
localization and 316
private 267
using locales to deteremine
URLs 317
attributes
acceptance of unquoted
295
allow-body 219–220
allow-informal-parameters
219, 221
class 219
copying 275
direction 224, 226, 229, 251
href 13, 250
HttpSession 272, 274
initial-value 308
key 314
of parameter element 222
quoting 296
serializing 275
session 22
src 230, 236
usemap 246
authentication 10, 96, 100,
407
automatic integration 339
AWT 18
B
bad coding shortcuts 15
banner ads
applications, building 337
BannerAd component 329
CreditCardField script
See CreditCardField script
data access 328
data model 326
DirectLink component of
325
function names 351
library specifications 336
page scripts 339
request servicing 334
script blocks 352
service implementation 332
service naming 335
source 327
storage 338
symbols 349
system 324
BaseComponent 35
BaseEngine 407, 416
BasePage 35, 57, 61
binary output 163
BinaryDumpOutputStream
163, 165
binding objects
See bindings
binding property 223
bindings
accessing 223
digit parameter 232
invariant 226
literal values 226
localized messages 226
plugged into component
parameters 223
storing for parameters 225
types of 223
updating values 226
blank fields 174
bodies, component
composition 220
discarding 220
Book 453, 457–458
Bookmarkable links 450
BookQuery bean 405–406, 419,
423, 425, 429–430
bookQuery property 419
bridging between stateless and
stateful worlds 299
browser back button 129, 132,
134, 148, 151, 154
bugs 271, 276, 308
business processes 44, 104
C
caching 52, 406
calculations in pages 234
Cascading Style Sheets (CSS)
97, 176, 184, 191, 201–202,
205, 217, 221, 261, 314,
511
changes, tracking 301–302
characters
invalid 292
non-ASCII, conversion of
292
nonprinting, conversion of
292
reserved, conversion of 292
unsafe, converting to HTML
entities 314
check boxes 104
choke points 14
ClassCastException 59
classes
AbstractComponent 218,
250, 304, 315
AbstractFormComponent
355, 363
AbstractService 334–335
Banner 327
BannerAd 329
BannerService 332
INDEX
classes (continued)
base 315, 335, 355, 363
BaseComponent 217–219,
302
BasePage 297
BaseValidator 198, 204
ComponentAddress 306
CreditCardField 348, 355–
356
CSS 221, 260
default 219
DirectLink 250
enhanced subclasses 308
examples.resume.Exerience
260
field-error 201
IActionListener 248
IEngineService instance 250
ILink 250
instantiation of a page’s Java
302
IValidationDelegate 208, 264
Java 179, 316, 356
java.io.PrintWriter 292
java.text.MessageFormt 316
label-error 201
Letter 237
ListEditMap 159
page 178
PatternValidator 199
PriorityModel 141
Register 194
RegisterDelegate 191
Tapestry 251
tapestry-palette 260
ToDoItem 137, 145, 155
UploadResults 163
utility 306
classpath 254, 256, 261, 267
client-side
JavaScript 174, 176, 198,
203, 339, 460
logic 6, 339
pop-up windows 203, 205
scripting 198, 203, 205, 339
validations 174, 176, 205
close tags, omitting 221
clustering 275, 308, 409
code
refactoring tools for changes
in 280
URL construction 280
URL servlet 280
code duplication removal 244–
245, 262
Collection 472, 500
complex form 105, 120
component libraries 480
application specifications 254
classpath location 264
classpath location of JAR 254
component referencing 256
component storage in classpath 261
Contrib library 254, 256
declaring use of 254
distribution as JAR files 254,
266
ID prefix 260
IDs 256
Java package for 264
naming patterns 256
packaging reusable components into 261
Palette component 256
prefixing 256, 260
secondary 254
servlets of 255
specifications 218, 262
tapestry-contrib-3.0.jar 254
using components from 263
component objects
framework 50
hierarchy 50
model 17, 478
component parameters 58, 80
component specifications 28,
78, 521
attributes and elements of
219
beginning of 219
instantiation of components
219
name of file 218
new property declarations
308
processing requests and 270
purpose of 219
sources of files 218
storage of 217
component templates
request processing and 270
539
component-oriented services
324
components 17, 21, 27
action triggering by 284
ActionLink 494
advanced form 134
anonymous Letter 237
Any 494
application-specific 216,
254
BannerAd 329, 339
Block 463, 495
Body 55, 167, 186, 220, 243,
344, 365, 495
body allowing 242
Border 217, 243, 258, 370
bound properties, reading
and updating 222
buffering the output of 345
Button 495
calculations during render
234
Checkbox 103, 110–111, 118,
121, 495
communication between 271
Conditional 98, 186, 210,
220, 496
configuration 75, 85
conflicts with informal
parameters 230
consistency between pages
and 299
construction details 290
contained vs enclosed 243
controller 240
controlling behavior of 222
converting application-specific to library components 264
corresponding to engine services 281
CreditCardField 341, 346,
348, 352
DatePicker 134, 167, 170,
208, 210–211, 221, 496
declared 259, 321
declared vs implicit 188
Delegator 165, 186, 221,
497
Digit
See Digit component
540
INDEX
components (continued)
Direct Area
See DirectArea component
DirectLink 55, 58–59, 62, 83,
86, 88, 98, 119, 222, 229,
237, 246, 281, 284–285,
288–289, 325, 333, 370,
497
duplications 244, 424
enclosed vs contained 243
end tags 220
ExceptionDisplay 498
exceptions and 221
ExternalLink 290, 450, 499
FieldLabel 173, 175, 181,
184, 188, 201, 208, 499
FileUpload 165
filtering benefits 295
final initializations 303
finishLoad() method, overriding 304
Foreach 79, 82–83, 85, 106,
115, 120, 124, 153–154,
156, 220, 222, 237, 239–
240, 500
Form 98–100, 102, 110, 149,
178, 181, 184, 188, 284–
285, 288, 299, 350, 361,
500
form element 355
FormError 264
Frame 501
GenericLink 501
Hidden 103, 134, 149, 153,
459, 502
IDs 188, 285–286, 289
Image 68, 73–74, 77–79, 221,
230, 233, 238–239, 266,
502
ImageSubmit 103, 110, 119,
316, 503
implementing in code 234
implicit components contained in templates 302
informal parameters 221
inputFirstName 194
inputLanguages 260
Insert 31, 187, 222, 292, 294,
314, 504
InsertText 504
instantiation of 219
interactive 246, 253
Java class of 217
JavaScript-enabled 340
JButton 21
Letter
See Letter components
libraries
See component libraries
LinkSubmit 505
ListEdit 103, 134, 151, 159,
464, 506
methods of
See methods
object graphs 253
Option 103, 110, 113, 135,
506
output-only 246
overview 134
page loader instantiation of
303
PageLink 225, 281, 333, 507
pages compared to 222
Palette 118, 256, 339, 398,
463
ParameterManager 226–227
parameters
See parameters
problems using sophisticated 259
problems with reusable 254
properties specific to pages
308
PropertySelection 103, 113,
118, 144, 210, 258–260,
318, 416, 507
purpose of 216
Radio 103, 110–111, 508
RadioGroup 103, 110–111,
508
RenderBlock 509
RenderBody 243, 434, 509
rendering 225
reusable 216, 246, 256, 265
reusable, packaging into
libraries 261
rewinding 228
Rollover 435, 457, 509
sameness of 320
Scaffold
See Scaffold component
Script 460, 510
Select 103, 110, 113, 135,
510
select 237
selectLoop 237
ServiceLink 511
Shell 97–98, 184–185, 243,
314, 511
simple 216
sources of 254
specifications
See component specifications
Spell
See Spell component
start tags 220
Submit 103, 110, 119, 121,
512
template
See template components
TextArea 102, 110, 210,
513
TextField 31, 98, 102, 107,
110, 121, 139, 149, 170,
184, 210, 222, 228, 513
Upload 103, 134, 160, 514
used in two different pages
308
ValidField 102, 293, 340, 367,
447, 514
See ValidField components
WEB-INF folder and 217
concurrent users 15
configuration files 12
connected parameter property
224
consistency 208
consistency goal 105
constraints 20
container 17
container-managed persistence 102, 404
context root 54
Contrib (Contributions) library
254, 256
Controller 29, 31, 33, 82, 117,
146
conversions 108
string 197
validation 171
value 173, 197
copyrights 243
INDEX
CreditCardField script
adding client-side JavaScript
365
blocks 352
body 352
Body component 344
component defining 341
CreditCardField component
348, 355–356
event handler functions 353
executing script specifications 365
form submissions 366
obtaining script specifications 365
script execution 365
symbols 349
CSS (Cascading Style Sheets)
See Cascading Style Sheets
(CSS)
cut-and-paste 53
D
data integrity 203
database connection leaks 15
database connection pools 15
database layer 19
database objects 136
database query 419
databases 5, 8, 15, 19–20, 30–
31, 44, 253
dates, formatting of 309
DateValidator 174
debugging 271, 276
declared component 55, 84
decoupling 49
Delegator 187
Demo applications 12
deployment descriptor 7, 14,
23, 28
descending property 423
design 16
desktop applications 4, 13, 16,
18, 93
developers
HTML 13–14
Java 13–14
Digit components
codeless 231, 234
digit parameter of 232
Guess and Win pages and 230
in parameter’s direction 232
input and output 230
specification 231
template 232–233
use of 233
direct services 250–251, 450
DirectLink components and
284–285, 288–289
engine services 284
Form components and 284–
285, 288
forms 288
listener methods, linking to
284
overview 284
request processing 287
stale session checking 285
URLs created by 290
DirectArea components
bodyless 248, 251
classes for 250
direct service and 250–251
DirectLink component similarity to 247
IDirect interface and 250
images and 252
implementation 248
Java class for 248
methods for 248, 250–251
parameters for 247–248
specifications for 248
static HTML, use of 252
support for informal parameters 247, 250
using 252
direction
auto 432, 434, 484
in 432, 454
direction attributes 434
directories
context root 371
specifications 264
disabled parameters 87
DOCTYPE 53, 475, 517
Document Object Model 462
document type definition 517
domain logic 20
domain object properties 104,
108, 355
541
domain objects 31, 44, 64, 96,
132
drop-down lists 113, 136, 144,
148, 318
DTD (Document Type Definition) 53, 295
duplications
removal of 244
unnecessary object 302
dynamic name generation 344
E
EAR (Enterprise Application
Archive) 384
Eclipse 35
Eclipse IDE 280
e-commerce applications 290
e-commerce sites, internationalization of 309
efficiency 52
EJBs (Enterprise JavaBeans)
384
elements
acceptance of unclosed 295
adding attributes to 240
application 256
area 246, 248, 250, 252
binding 233, 349
body 352
component 219, 259
component-specification
219–220
context-asset 219
discarding from templates
245
dynamic 252
dynamic tags, checking validity of 295
img 230, 236–237, 240, 246
initialization 354
input-symbol 348
let 348–349, 351, 366
library 256
map 246, 252
matching start and end tags
295–296
parameter 219, 222, 224, 251
private-asset 267
property-specification 219,
306, 308
542
INDEX
elements (continued)
rendering vs previewing conflict 245
reserved-parameter 219, 230
script 294, 353
select 257–258, 261
set 348–349, 366
span 244, 314
specification 267
static 252
table 258, 260, 265
unique 352–353
See also HTML elements
engine 406
engine locale 317
engine service 530
Engine services 246
engine services
application operations 281
application URLs 279, 284
components 323
creating new 323
delegation layer 277
direct service 284
external service 290
home service 281
implementing 332
library for 336
naming 335
page rendering 281
page service 283
services roster 281
servlets compared to 281
URL construction 372
engineered coincidence 98
Enterprise Application Archive
(EAR) 384
enterprise application deployment 477
enterprise applications 12, 476
Enterprise JavaBeans (EJBs)
19–20, 23, 384, 404, 407
entity beans 404, 407, 410
entity objects 44
Enum base class 144
EnumPropertySelectionModel
147
Enums 148, 208
errors
handling 187
identification of 280
messages 156, 192, 267, 315
properties 416, 418, 465
event handlers 203, 221
connecting to objects 353
onkeyup 343–344
onsubmit 344, 351, 355, 366
window.onload 344
event notification 21, 117
event registration, automatic 303
example applications
Hangman application
See Hangman application
Java Pet Store
See Java Pet Store application
Resume Builder
See Resume Builder application
examples (servlet name) 255
examples.register.RegisterDelegate 191
Exception page 109
exception report 23, 58
ExceptionAnalyzer 498
ExceptionDescription 498
exceptions 221, 299
RedirectException 370
ValidatorException 368
expired session 414
external services 290, 450–451
Externalizable 410
Extreme Programming 481
F
fabricated subclass 101
fail-over strategy 274
field decorations 175, 177, 186,
189, 199, 202, 208, 210
file accessibility for users 266
file extensions
.application 254
.jwc 217
file uploads 134, 165, 514
filtering
HTML entity, disabling use of
315
output 292
firewalls 274
form control components 95,
98, 103, 108, 110, 113
form control elements 94, 104
form controls 228
form element ids 105, 113
form submissions 12, 98, 104,
107–109, 112, 194
formal parameters 527
format 504
form-level validation errors
207
form-level validations 208, 263
form-related components,
tracking 210
forms 22
complex 105, 120
components
See components
delegate 362
direct service and 288
loops 159
rewinding 152, 288, 363
simple 105
submissions 271, 366, 371,
375
synchronization issues 153
frameworks 16
commons-lang 144
JavaBeans 17, 101
JUnit 482
Maverick 11
OGNL 31
OGNL library 144
Struts 11–12
Turbine 11
Velocity 12
WebObjects 18
WebWork 11
functions
anonymous JavaScript 354
onkeyup_cc_field() 354
G
Game 45, 47, 61–62, 64, 79
Global object 406
goals
consistency 22, 60, 89, 100,
105, 136, 292
efficiency 23, 52
feedback 23, 60
simplicity 60, 100, 105
graphs, complex object 302
INDEX
Guess page 42, 65, 230, 233–
234, 237–238, 240–241
GUI 18, 21, 29
H
Hangman application
Body component 243
Border component 243
changes in original code 237,
241
components with bodies 242
Guess page 230, 233–234,
237–238, 240–241
Letter component 234
Lose page 234, 238
overview 39
RenderBody component 243
rendering vs previewing conflicts 245
review of previous 216
revised compared to original
217
Shell component 243
source code online 217
Spell component 238
stylesheet 244
target word 238, 240
Win page 230, 234, 238
HashSet 117
helper beans 179, 187, 189,
191, 195, 212, 518
hidden fields 107, 131
hidden form fields 134, 149,
151
home interfaces 23, 406
Home page 50–51, 54, 281
home service 281
hot code replace 405
href 13
HTML
editors
See WYSIWYG HTML editors
safe 292
static mockups 342
HTML developers 13–14
HTML elements
a 13, 55, 87, 434
bean 178
body 167, 185, 415
form 13, 94, 98, 102–105
head 97–98, 184–185, 511
img 68, 73
input 94, 103, 107, 149, 160,
173, 177, 197–198, 201
link 184
option 103, 113, 135, 140,
506
script 167
select 94, 103, 113, 135, 140,
510
span 68, 80, 124, 176, 201
table 124, 187, 413
textarea 94, 102, 110
title 97–98
tr 124, 131, 425
See also elements
HTML mockups 68, 75, 83
HTML templates 51, 54, 234
HTTP 6–7, 11, 15, 18–19, 64
as stateful protocol, maintaining illusion of 306
bridging between servlets and
299
protocol 271
HTTP cookies 11, 15
HttpServlet 21, 25
HttpServletContext 25
HttpServletRequest 7, 9–12,
15–16, 21, 25, 30, 35, 94,
104
HttpServletResponse 35
HttpSession 11, 15–16, 21, 25,
35, 60–61, 64, 126, 128,
149, 406, 409, 414, 498
HttpSession attribute 60
HttpSessions, expired 251
Hypertext Transport Protocol 6
I
IActionListener 59
IActivate 443, 445
IActive 411
IAsset 75, 503, 510
ICallback 442
IComponent 33, 35, 454
IDEs (integrated development
environments) 13, 280
Eclipse 405
543
IDs, unique 155
IEngine 35
IErrorProperty 411, 416, 418
IExternalPage 411, 451, 499
ILibrarySpecification 523
image maps 246
images
assets 237
client-side JavaScript and
256
Letter component 237
localization of 270, 311, 316
manipulation 217
problems with icon 266
referencing packaged 266
serving from the JAR file
266
using assets to represent 316
IMarkupWriter 35
IMessageProperty 411, 416,
418
implicit component 55
informal parameters 78, 221,
520, 530
initializations, performing 299
input symbols 510
input validation 104
InsertTextMode 505
instance variables 9, 101
instrumentation 43
Instrumenting 50
integers, compared to enums
144
integrated development environments (IDEs) 280, 405
interactive pages 246
interfaces 17
event 303
HTMLWriter 297
HttpServletResponse 297
IActionListener 250
IBanner 327, 336
IBannerSource 327, 331,
336, 338
IBinding 223
IComponent 288, 292, 304
IDirect 250, 287
IEngineService 278–279, 281
IExternalPage 290
IForm 361
ILink 333
544
INDEX
interfaces (continued)
IMarkupWriter 201, 237,
292, 297
IPage 290, 299
IPageLoaderpageloader 302
IPropertySelectionModel
260, 318
IRender 164–165, 187, 292
IRequestCycle 282, 300, 317
IValidationDelegate 191, 199,
202, 367
IValidator 173, 198, 204
PageDetachListener 303,
305–306
PageRedirectException 283
PageRenderListener 297,
303
PageValidateListener 283,
303
internationalized 4, 15
Internet Explorer 6
invalid input, handling 170
IPage 33, 35, 61
IPropertySelectionModel 135,
140, 142, 147, 416
IRender 33, 497–498
IRequestCycle 35, 58, 88, 410,
442
Iterator 472, 500, 506
IUploadFile 160, 163, 165, 514
IValidator 515
J
J2EE 404
Jakarta 12
Jakarta commons-fileupload
library 160
Jakarta commons-lang framework 144
Jakarta ORO 199
Jakarta Tomcat 476
JAR (Java Archive) files 27, 254,
266
Java 2 Enterprise Edition
(J2EE) 246, 274, 404
Java class default 217
Java classpath 254, 256
Java Database Connectivity 20,
405
Java developers 13–14
Java Naming and Directory
Interface 23, 406
Java Pet Store application
DirectArea component 247
DirectLink component 246
image map 246
listener methods 246, 248
pet inventory 253
PetCategory page 252
Pets page 252
Java primitive types 228, 232
Java Runtime Environment
(JRE), localizing applications with 309
Java Servlet API 7, 16, 25, 37
Java Standard Tag Library 33
Java types
array 228
Boolean 228
char 240
double 228
int 228, 232
java.lang.Object 248
Java Virtual Machine 405, 410,
429
java.io.PrintWriter 201
JavaBeans 12, 17, 33, 188, 222–
224
application servlets and 271
framework 101
instantiation 178
properties 47, 87, 101, 459,
478
JavaScript 20, 495, 505, 532
anonymous functions 354
Body component and 344
client-side 134, 165, 167,
172, 174, 186, 198, 203,
256, 339
components enabled by 340
customized on demand 340
DOM references 350
dynamic generation 340
event handlers 221, 344, 352
filtering 294
functions 351
malicious 294
JavaScript library 534
JavaServer Pages (JSPs) 7, 279
JBoss 7, 18, 385
JDBC 419–420
Jetty 25, 482
JRE (Java Runtime Environment), localizing applications with 309
JSP tags 78, 88, 280
JSPForm page 377
JSPs (JavaServer Pages) 7–15,
21, 30–31, 37, 43, 52, 63–
64, 77–78, 87–88, 95, 279
custom tags 372, 374–378
ID component determination 376
integrating with Tapestry
applications 369
linking back to applications
371
redirecting application
requests to 369
response rendering with 369
JSTL 33
JUnit framework 482
jwcid attribute 55, 68, 84, 295
$content$ 245
$remove$ 245
Jython 525
K
key parameters 315, 317
KeyAllocator bean 405
keys, primary/unique 155
L
label decorations 175, 188, 202
language proficiency 263
languages
handling differences in 292
handling multiple, using JRE
310
locales compared to 309
translation of 309
layout controls 258
Letter components
accessing connected parameter properties 237
code 236
components used in 237
correct asset for 236
Guess page and 237
implementation 236
INDEX
Letter components (continued)
methods for 237
parameters 235–237
purpose of 234
renderComponent() method,
use of 234
similarity to Digit component 235
simplicity of 238
simplification of template
238
specifications 235
template free 234
using 237
libraries
See component libraries
library specifications 525
lifecycles
page 179, 191–192
request 179, 191
light touch 91
links
bookmarkable 290
method invoking 253
List 126, 128, 472, 506
ListEditMap 464, 468, 470
listener methods 22, 33, 55, 58–
59, 62, 73, 86, 88, 96, 98–
100, 109, 119, 127, 195,
208, 210, 252, 419, 437
as class instance 248
BaseComponent and 220
calling back application-specific code 253
defining for Java Pet Store
246
forms 317
IActionListener 149
invoking 151–152, 222, 250,
288
JSP and 369
linking to, with direct service
284
synchronizeItem() 156, 159
listener objects 250
listener parameters 58
listeners
adding and removing 299
non-null 251
property 59
lists 151, 153
literals
localized 321
simple 314
load balancing 274
locales
current, passing to model’s
constructor 318
definition 309
engine, editing 317
languages compared to 309
matching page and engine
317
self-describing 320
used by assets 317
localization
defined 309
handling with JRE 310
independence of page 320
logic
application-specific 57
LoginException 449
look and feel of pages 42, 50,
65, 79
Lose page 62, 234, 238
M
Map 409, 458
Map components 188
markup languages, types of 291
markup writer, calls to 250
markup writers 292
MasterQueryParameters 429
Maverick 11
messages
assembling from pieces 315
localized 223, 226, 313
properties 416, 418, 465
specialized error 315
warning, localized 315
metadata 417, 529
methods
accessor 223, 225, 228, 237
actionTriggered() 250–251,
288
activate() 62, 88, 282, 288,
442, 445
activateExternalPage() 290–
291, 378, 451, 499
addEventHandler() 361, 366
addPageRenderListener() 299
545
attach() 301
attribute() 237
beginEmpty() 237
beginResponse() 297
checkExtension() 331
checkRequired() 198
cleanUp() 305
cleanupAfterRender() 226
commitPageChanges() 297
constructLink() 334
constructServiceParameter()
250
createNewItem() 139, 146,
159
detach() 305–306
doPost() 108
encodeURL() 334
end() 201
equals() 145
event 306, 335
extractCardNumber() 364
finishLoad() 226, 303, 308,
428, 441
fireObservedChange() 307–
308
forgetPage() 449
format() 315
GET 377
get 237
getAsset() 74, 77
getBeans() 195
getContentType() 296–297
getDisplayLanguage() 320
getExtension() 331
getForm() 361
getKeys() 159
getLabel() 142
getLink() 333
getListener() 251
getLocale() 147
getMessage() 313, 315
getName() 335
getObject() 223–224, 228
getOption() 142–143
getOptionCount() 142
getPage() 282, 300–301, 303,
317
getPriorityModel() 139
getRandomBanner() 331
getResponseWriter() 297
getServiceParameters() 88
546
INDEX
methods (continued)
getSession() 11
getStream() 165
getTargetPage 225
getTaskWarning() 316
getValue() 142, 158
handleStaleSessionException() 414
initialize() 305–306, 332, 334
isInError() 201
isStateful() 248, 250–251
links that invoke 253
match() 199
pageBeginRender() 117, 126,
158, 194, 288, 297, 432,
446, 452, 456, 459, 467
pageDetached() 306
pageEndRender() 288, 297
pageValidate() 335, 444
POST 377
prepareForRender() 225
print() 292
record() 207
recordFieldInputValue() 367
removePageRenderListener() 299
render() 165, 187, 225, 292,
297–298, 497
renderBody() 56, 251
renderComponent() 34, 56,
225, 227, 230, 234, 237,
248, 250–251, 331, 335,
363, 365, 433
renderInformalParameters()
250
renderPage() 297
renderResponse() 283, 297
renderValidatorContribution() 176, 204
resetParameters() 226
service 273
service() 9, 272, 277, 334
set 237
setAttribute() 276
setContentType() 297
setFormComponent() 207
setObject() 223–224, 228
setParameters() 226
setter 307–308
static 307
static utility 250
synchronizeItem() 159
toObject() 199
toString() 197
translateValue() 142
trigger() 248, 250–251, 288
updateCardNumberFromRequest() 366, 368
updateSet() 472
validate() 282, 284, 287–288
writeAttributes() 176, 202
writeLabelPrefix() 176, 201
writeLabelSuffix() 176, 201
writePrefix() 202
writeSuffix() 202
MIMEs (Multipurpose Internet
Mail Extensions) 160, 296–
297
Model 29, 47, 49, 75, 79, 82,
117
Model-Driven Architecture
529
Model-View-Controller pattern
29, 39, 44, 47, 79, 91, 135,
146, 478
multipart/form-data 160–161
multiple selections 113, 115,
118, 135
multi-threading 9, 12, 16,
271
N
names
conflicts 344
reserved 230
namespace conflicts 279
naming conventions 17, 101
navigation menus 415
navigational controls 243
Netscape Navigator 6
news sites, internationalization
of 309
NeXT 18
NoSuchPropertyException 59
null checking 194
null value translation 198
null values 208
NullPointerException 117
numbers, formatting of 309
NumberValidator 174
O
object relational database 18
objects
applications servlets 270
database 136
Date 198
engine services 270
engines 270, 272, 276–277
for request processing 270
Game 239
graph 32
HttpSession 275, 285
immutable 276
Integer 198
mutable 276–277
Order 136
Page Loader 270
Page Recorder 270
Page Source 270
page, pooling instances of
272
Pattern 199
pools 272, 300
renderable 187
request cycle 270
RequestContext 270, 273
serialization and deserialization 145, 406
serializing for storage 302
setting properties for 297
unnecessary duplication of
302
Visit 239, 272, 277
OGNL expressions 31, 33, 37,
58–59, 74, 77, 84, 98, 423,
519, 529
beans.delegate 187–188,
191
beans.required 179
bindings based upon 226
complex parameter bindings
223
components.inputFirstName
188
creating symbols with 348
default values 260
digit properties and 232
evaluating 308, 349
getMessage() method as part
of 313
INDEX
OGNL expressions (continued)
locating in the body 193
page.visit.game.letters 239
parameters that are 313
placed in the element body
233
updating 223
values for, used to set and
update properties 308
open source 12
operation signatures 14
Operations bean 405–407, 416,
449, 458, 464, 468
org.apache.tapestry.visit-class 90
output
compatability with XML style
markup 296
filtered 293
streamlining markup 292
writing directly within Java
code 292
P
page
class 50, 52–53, 57
instance 52
loaders 302
mockup 42, 44, 50
properties 60, 111
recorders 297, 301
rendering 56
services 283
sources 301, 317
templates 220, 270
page specifications 28, 49–51,
55, 76–77, 84, 527
caching 270, 299
new property declarations
308
non-localization of 321
parsing 270
PageCallback 442, 445, 453
page-oriented services 324
PageRedirectException 159,
445, 467
PageRenderListener 117, 126,
432, 446, 467
pages
active, inability to change
locale of 317
as controllers 240
attachment to application
engine 301
caching by request cycle
object 300
cleaning prior to storage
305
cleanup 306
components compared to
222
consistency between components and 299
creating new instances 302
division between instances
and persistent page state
302
Exception 109
final initializations 303
finishLoad() method,overriding 304
Guess 42, 65, 230, 234, 237–
238, 240–241
Home 50–51, 54, 281
HTML rendered, compared
to XML related 296
independence of localizations 320
instances, construction of
299
instances, readying 300
issues in using stateful 300
locales
See locales
localized strings and static
portion of 311
localizing text and images
270, 292
Lose 62, 234, 238
non-validation of 295
obtaining fresh instances of
301
page loader and 303
page source 301
persistent page properties
302, 306
persistent state separate from
instance of 301
pooling instances of 270
pools 299
previewing 230
properties files 313
547
recorders 297, 301–302, 307
recursive processing 302
references, problems with
storing 306
rendering sequence 297
renderingpage 281, 291
resetting properties of, to
pristine values 304
returning to pools 304
reusing 300
rewinding 299, 363
serializing and deserializing
302
servlet use, compared to use
of 300
specifications
See page specifications
specified properties, supporting 308
StaleLink 109, 129
StaleSession 286, 414
storage of names 306
storage of persistent properties 302
stylesheets 260–261
supporting objects associated
with 302
using multiple templates for
320
validation with external service 291
variable target 371
Win 62, 230, 234, 238
WYSIWYGs and 230
PageValidateListener 444
Parameter 27
parameter direction 528
ParameterManager 226–227
parameters 87
AbstractComponent 225
accessing 223
accessing values bound to 225
alt 314
analyzing direction of 226
application-specific service
286
attributes, converting from
informal 251
binding 58
binding objects 222
See also bindings
548
INDEX
parameters (continued)
binding property names and
names of 223
boolean 355
cardNumber 356, 368
component 201
condition 99
connected parameter property and 225
converting to objects array
250
cycle 292
delegate 165, 178, 181, 184,
187, 266
digit 232, 235
direction of, choosing 229
direction values of 226, 483,
528
DirectLink 221
disabled 87, 235, 237, 356–
357, 436
displayName 173, 191, 193,
201, 209
element 124
field 173, 188
focus 436
forbidding informal 221
forbidding names 230
form 226
formal 191, 222–223, 230,
314
hidden 98
image 230, 436
in 226, 251
informal 191–192, 221, 230,
314
JavaBean objects compared to
component 223
key 315
letter 235–237
listener 58, 98, 119, 152, 222,
229, 247–248, 251
model 135
multiple 115
naming conflicts 225
non-rendering components
and 229
page 225, 290
ParameterManager 227
parameters 87, 247–248,
250, 290
prefixes used to identify 313
processing of 225
properties 228, 237
purpose of 222
query 228, 279, 282, 284,
286, 377–378
reading and converting 224
read-only 229
reserved names 230
reusable components and
216
rewinding 224
selected 110–112, 119, 127,
260
service 246, 285, 290, 377–
378
service query 284
setting properties 225
setting values 226
size 192
sort 260
source 151–153, 222
stateful 285, 414
stylesheet 184
supporting 308
tag 119, 127
title 314
unknown runtime types and
248
updating 223
userName 378
ValidationConstraint 208
validator 174, 178, 191, 193
value 110, 112, 139, 142,
149, 151–153, 191, 193,
197, 199, 222, 228
when majority are used 229
writer 292
parsing specifications 52
password field 98
patterns
Model-View-Controller 355
performance 52, 78
performance enhancement 406
performance profiling 410
Perl 6
persistent page properties 60,
101, 125, 149, 301–302,
306, 432, 484
Person 409–410
Person bean 407, 410
placeholders 315
plumbing 178
pooling instances of page
objects 270, 272
pools
database connection 15, 272,
300
object 272
obtaining page instances from
317
page 299
pop-up calendars 165
pop-up windows 203, 205
prefixes
tapestry 374
prefixing 256, 279, 314
presentation layer 19–20, 275,
404, 410–411, 449
previewing pages 238, 240–241,
245, 314
primary key 407
primitive Java types 228, 232
priority levels, selecting 141
PriorityStrings.properties file
147
processing requests
See request processing
properties
abstract 356
address 195
assets 184
beans 179
cardNumber 356
clientScriptingEnabled 204
components 181
connected parameter 225,
237
contentRenderer 165
currentDate 315
danger of leaving in place 304
Date 174
date 167
declared, initial values of 308
digit 232
dueDate 315
endDate 205, 207
engine 305
errorMessage 156
expressions 31
firstError 187
firstName 194
INDEX
properties (continued)
form 356
game 239
hasErrors 186–187
inError 209–210
item 156, 159
JavaBean 12, 223
key 153, 158
languages 260
letter 240
listEditMap 156
listeners 59
locale 147, 317
name 33, 356
optimizing number of 226
page 225, 239
persistent page 270, 277,
292, 297, 301
priority 138–139
priorityModel 139, 146
reading and updating bound
222
reinitializing 305
rendering 224
requestCycle 305
required 174, 192, 198
setting for objects 297
setting to null 305
startDate 205, 207
state 192
synthetic 72
targetPage 225
toDoList 156, 158
tracking changes to persistent
page 301–302
updating 224, 228
validity of 228
visit 239, 305
Protected 411, 445
prototyping 75
public static field 423
Publisher 464
Python 525
Q
query parameters 5, 7–9, 12,
14, 16, 22, 37, 63, 86, 94,
98, 100, 104–105, 110, 118,
160, 279, 282, 284, 286
R
race conditions 159
radio buttons 104
radio elements 111
recompile 280
recursive processing 302
redeploy 406
redeployment 407
RedirectFilter 475
refactoring tools 280
Register page
building 189
HTML markup and 293
implementation 195
specifications 195
RegisterDelegate 202
registration, automatic 299
remote interface 407, 425, 429
Remote Method Invocation
405
remote object access 407
RemoteException 406, 426,
429–430, 432
render events 288
request cycle rewinding 250
request cycles 6, 8, 16, 35, 270,
292, 297, 300, 304
request processing
attribute serializing and copying 275
direct service 287
end of 304
external service 290
initial access of a page 317
order operations 287
and 306
request cycles 292, 297, 300,
304
server-side 339
Servlet API and 276
stateful 286
synchronization of serverside-state 276
RequestContext 498
reserved entities, escaping 296
Resin 25
resource bundles 310
ResourceBundle 148
resources, managing scarce 272
549
Resume Builder
form-level validation 263
Palette component 257
stylesheet 260–261
retry loops 407, 429, 432, 470
reuse 23
rewind phase 109, 126–127,
129
routers 274
S
Scaffold component
choice of images 231
codeless 231, 234
digit parameter of 232
in parameter’s direction 232
specification 231
template 233
use of 234
script specification elements
460
body 462, 533
foreach 533
if 534
if-not 534
include-script 534
initialization 462, 534
input-symbol 535
let 535
set 535
unique 536
script specifications 532
scripting languages 6
scripting, client-side 198, 203,
205, 339
scriptlets 77, 87–88
scripts
blocks 352
security 203
checks 282, 284
filtering and 293–294
separation of concerns 29, 76
separation of layers 405
Serializable 410
serialVersionUID 410
server clusters
request processing with 274,
306
550
INDEX
servers
JBoss application 385
open-source 385
server-side
redirects to the JSP 376
values, converting 364
server-side state 10–12, 18, 35,
60–61, 101, 109, 131, 270,
272, 302
server-side state management 18
service parameters 88
Servlet API 271–273, 275, 292
Servlet API and multipart/formdata 160
ServletContext 406
servlets 5, 7, 11–13, 15, 22, 27,
30, 37, 63, 86, 88–89
applications 340
classes 64
containers 7, 9, 11, 18–19,
25, 28, 474–475, 482
context 90
filter 26, 475
names 255
See also application servlets
session attributes 22, 409
session beans 404
BookQuery 405–406, 419,
423, 425, 429–430
KeyAllocator 405
Operations 405–407, 416,
449, 458, 464, 468
sessions 11
Set 117, 472
shopping carts 151
simple forms 105
simplicity goal 22, 105
single selections 115
singletons
data 328
SortColumn 423
sortColumn property 423
special component ids
$content$ 245
$remove$ 83
specification elements
application 517
bean 518
binding 84, 191, 519
component 70, 84, 520
component-specification 521
component-type 521
configure 520
context-asset 70, 75, 523
description 523
duplication removal 244, 262
extension 523
external-asset 524
inherited-binding 524
library 525
library-specification 525
listener-binding 525
page 526
page-specification 527
parameter 483, 527
parsing 52
private-asset 528
property 417, 529
property-specification 102,
417–418, 529
reserved-parameter 530
service 530
set-property 179, 193, 531
set-string-property 193
static-binding 191, 531
specified properties 100, 156
Spell component
basis for 239
binding letter property to letter parameter 240
bodyless 241
components for 239–240
Guess page and 240–241
implementation 239
properties of 239
purpose of 239
replacing with spaceholder
241
simplifying page template
with a 239
target word and 240
updating the letter property
240
using 240
Spindle 35–36
stack trace 23
stale links 129
stale sessions 285, 414
StaleLink page 109
StaleLinkException 120, 129,
131, 151, 154
StaleSession pages 414
StaleSessionException 414
stateful flags 285
stateful session beans 419, 430
stateless applications 61
stateless coding 272
stateless mode 61
stateless protocol 6
stateless session beans 407
states
application-wide 277
internal 277
server-side 270, 274
stateValidator 192
static HTML 220, 252
static text 220, 296
storage
application-wide states 277
assets 267
cleaning pages prior to 305
component specifications 254
engine instances in HttpSession 272
global application data 272
in package directorie 264
object states 302
page instances 317
page names 306
page specifications 254
302
serializing objects for 302
specifications 261
WEB-INF folders 254
strings
compiling 199
conversion of 174
editing properties of 171
empty 198
key parameters 315
literal 193
localization of, using JRE 311
localization of, using Tapestry 311, 314
message, operating with missing 315
static (literal) 191
StringValidator 174, 191–192
Struts 12, 37, 98
Action 12
INDEX
stylesheets 184, 258
submit controls 118
subsystems 35
suffixing 279
super class implementation,
invoking 304
Swing 18, 21, 29
symbols
baseName 350
creating 349
creditCardField 349
formatMessage 349
formValidateFunction 351,
361, 366
input 349, 365
Map 366
uniqueness, importance of
351
synchronization faults 151
synthetic properties 72, 82,
115, 135
system properties 25
templating 7, 12
test 91
test suite 49
testing 49
text
formatting 316
instructional, need for localization of 314
localization of 270, 311, 320
static 296
translation 309
text fields 104–105
thread of control 16
tier
application 44
title bars 243
ToDo application 159
Tomcat 18, 25, 39, 487
toolkit 16
Turbine 11
T
unit-test suite 481
unsafe characters, converting
237
unwanted dependencies 13,
98
updating applications 155
URLs (Uniform Resource Locators) 5, 13–14, 16, 22, 27,
55, 63–64, 86, 88, 90, 98
application 279, 284
application servlet 290
determining with assets 317
representation, Tapestry
applications 333
user identity 6, 96
UserListEditMap 471
utility functions
onkeyup_cc_field() 352
validate_cc_field() 352
Tacos
See Tapestry Component
Archive
tag libraries 12
tag library descriptors 374
taglib declarations 377
tags
See HTML elements
Tapestry Component Archive
480
team development 4, 64
template components
body discard 220
instantiation of class instance
217
page template composition
220
specifications 217
See also component specifications
steps in creating 217
template parser 295
templates
caching 299
multiple 320
U
V
validation 12
delegates 362, 364, 367
input subsystems 364
subsystems 367, 388
validation delegates 449
defining 193
551
field and error tracking with
173
helper beans and 179
label and field decoration and
202
listener methods and 207
ValidField components and
178
without ValidField 209
ValidationDelegate 177, 202
ValidationException 175
validations
client-side 170, 205
conversions 171
errorserrors 187
form-level 208
server-side 171, 202
ValidatorException 198
validators 283
creating 198
custom 199
defining 193
helper beans and 179
Register pages and 189
stateValidator 192
ValidField 203
ValidField components
client-side JavaScript produced by 186
FieldLabel components
mixed with 208
FieldLabel, using with 181
inputEnd 205
inputStart 205
integration of client-side validation into applications
203
JavaBeans andvalidfield
178
Register page specifications
and 194
TextField variation 170
validation delegates shared by
184
validation without 208
writeAttributes() and 202
writeSuffix() and 202
value conversions 173, 197
values
accessing bound 225
auto 228
552
INDEX
values (continued)
bound to component parameters 311
custom 228
direction of movement 224
form 228
in 228
initial, when none are provided 309
limitations 228
literal 226
literal static 223
OGNL expression default
260
ognl null 251
problems with mutable 307
properly quoting 237
reading and updating bound
223
reserved characters 292
service context 285
stateful flag 285
updating 226
variables
instance 271
Velocity 12
View 29–30, 75, 146
Virtual Library (VL)
BookLink component 422,
426, 450
BookMatches page 413, 418,
425, 430, 434
Border component 415,
438
Borrow component 426
BorrowedBooks page 430
Browser component 425,
430, 432
ColumnSorter component
423, 434
EditPublishers page 464
EditUsers page 470
GiveAwayBooks page 463
Home page 413
MyLibrary page 430, 444,
457
NewBook page 445
PersonLink component 422,
426, 453
ShowError component 416,
418
ShowMessage component
416, 418
ViewBook page 411, 422,
426, 449, 451, 453
ViewPerson page 422, 426,
449, 455
Virtual Library application
accessing 385
Add New Book page 396
administration
See VL administration
Ant Build script 385
book borrowing 389
Borrowed Books page 395
configuration 385
database access 384
database sample 389
database updating 389
defined 384
detail pages 390
Edit Book page 395
Edit Profile page 396
Edit Publishers page 400
Edit Users page 399
Enterprise Application
Archive EAR) 384
Enterprise JavaBeans (EJBs)
384
functionality, adminstrative
399
functionality, user 399
Give Away Books page 396
JBoss application server and
385
log in on the fly 392
logging in 388
Login page 388
See also VL Login page
managing books 393
Matching Books page
See VL Matching Books
page
My Library page
See VL My Library page
non-lendable books 393
Register page 389
registering 388
Search page
See VL Search page
Transfer Books page 400
View Book page 390
View Person page 392, 411,
430
VirtualLibraryEngine 409–410,
414, 430
Visit object 58, 60, 406, 409
VL administration
administrative privileges 399
book ownership transfer 399
Edit Publishers page 400
Edit Users page 399
form wizard 400
Give Away books page 400
navigation menu 400
pages, protection of 399
passwords 399
publisher list, managing 399
system lock-out 399
testing administrative pages
399
Transfer Books page 400
users, deleting 399
VL Edit Book page
drop-down blanks 395
drop-down lists 395
Publisher field 395
VL Login page
email addresses, unknown 388
error messages 388
fields, invalid 388
login ID 388
passwords 388
Register now link 389
return users 389
validation subsystems 388
VL Matching Books page
book borrowing 386, 389
column sorting 387
disabled button 387
links 386
navigation options 386
new books icon 386
non-lendable books 386
page navigation controls 387
paging results 387
sort order controls 387
table 386
VL My Library page
Add New Book page 396
books, deleting 393
books, editing 393
Borrowed Books page 395
INDEX
VL My Library page (continued)
Delete Book icon 395
deleted books confirmation
395
Edit Book page 394
Edit Profile page 396
navigation controls 394
returning books 395
VL Search page
form 385
navigation menu 385
navigational border 385
page title 385
reminder 391
Search button 386
VL View Book page 390
VL View Person page 392
W
WAR 27, 52, 75, 90, 476
warning messages, localized
315
weak binding 13, 95
web applications 4
context 27, 523
credit card numbers 341
localization of 309
web deployment descriptor 37,
61, 63–64, 88–89, 474–475
web sites, internationalization of
309
WEB-INF 52
WEB-INF folder 217, 254, 266,
313, 365
WEB-INF/lib folder 254
WebLogic 7, 18, 25
WebLogic application servers
276
WebObjects 18
WebSphere 7, 18, 25
Win page 62, 230, 234, 238
window titles 315
WML (Wireless Markup Language) 291, 295, 494
WordSource 45, 47, 61, 64
553
WYSIWYG HTML editors 14,
42, 44, 68, 73, 78, 83, 314
editing, support of 221
previewing with 230, 233,
241, 259
WYSIWYG preview 68, 184,
415
X
XHTML 291, 295
XML (Extensible Markup Language) 7, 19, 23, 35, 37,
291, 295
files 374
XML namespaces 256
XML-related markup languag

Tapestry and HTML forms

Transcription

Similar documents

Wall Tapestry

office - Nightwood Theatre

Deerhorn Proposal - Ciat

“Gawu” — African-influenced Tapestry

LCLPB OS ENG

2013-10 CommUUnique - Tapestry, a Unitarian Universalist

Rodrigo Santoro

Marina Terra 303